diff --git a/docs/1intro/case-study.ipynb b/docs/1intro/case-study.ipynb
index 4acacbf..7231915 100644
--- a/docs/1intro/case-study.ipynb
+++ b/docs/1intro/case-study.ipynb
@@ -17,7 +17,7 @@
     "\n",
     "1. Daten laden und verstehen\n",
     "\n",
-    "   * Nutzung eines offenen Datensatzes (z.B. Boston Housing Dataset oder Kaggle Immobilienpreise).\n",
+    "   * Nutzung eines offenen Datensatzes (z.B. California Housing Dataset oder Kaggle Immobilienpreise).\n",
     "   * Untersuchung der Datenverteilung, Korrelationen und möglicher Ausreißer.\n",
     "\n",
     "2. Datenvorbereitung\n",
@@ -39,6 +39,47 @@
     "## Code-Beispiel"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "d7f69c6d-4db0-484d-9c68-c3ec4fce16b9",
+   "metadata": {
+    "collapsed": true,
+    "jupyter": {
+     "outputs_hidden": true
+    }
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Defaulting to user installation because normal site-packages is not writeable\n",
+      "Collecting scikit-learn\n",
+      "  Downloading scikit_learn-1.6.1-cp39-cp39-macosx_12_0_arm64.whl.metadata (31 kB)\n",
+      "Requirement already satisfied: numpy>=1.19.5 in /Users/minye/Library/Python/3.9/lib/python/site-packages (from scikit-learn) (1.26.3)\n",
+      "Requirement already satisfied: scipy>=1.6.0 in /Users/minye/Library/Python/3.9/lib/python/site-packages (from scikit-learn) (1.12.0)\n",
+      "Collecting joblib>=1.2.0 (from scikit-learn)\n",
+      "  Using cached joblib-1.4.2-py3-none-any.whl.metadata (5.4 kB)\n",
+      "Collecting threadpoolctl>=3.1.0 (from scikit-learn)\n",
+      "  Using cached threadpoolctl-3.5.0-py3-none-any.whl.metadata (13 kB)\n",
+      "Downloading scikit_learn-1.6.1-cp39-cp39-macosx_12_0_arm64.whl (11.1 MB)\n",
+      "\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m11.1/11.1 MB\u001b[0m \u001b[31m6.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m00:01\u001b[0m0:01\u001b[0m\n",
+      "\u001b[?25hUsing cached joblib-1.4.2-py3-none-any.whl (301 kB)\n",
+      "Using cached threadpoolctl-3.5.0-py3-none-any.whl (18 kB)\n",
+      "Installing collected packages: threadpoolctl, joblib, scikit-learn\n",
+      "Successfully installed joblib-1.4.2 scikit-learn-1.6.1 threadpoolctl-3.5.0\n",
+      "\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip is available: \u001b[0m\u001b[31;49m24.3.1\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m25.0.1\u001b[0m\n",
+      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49m/Applications/Xcode.app/Contents/Developer/usr/bin/python3 -m pip install --upgrade pip\u001b[0m\n"
+     ]
+    }
+   ],
+   "source": [
+    "import sys\n",
+    "# !{sys.executable} -m pip install scikit-learn\n"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "9d22c329-4a99-4a23-af24-6e4b2eafec89",
@@ -67,7 +108,7 @@
    "id": "7346190a-b639-455f-95ef-420470149e57",
    "metadata": {},
    "source": [
-    "### Beispieldatensatz laden (Boston Housing Dataset)"
+    "### Beispieldatensatz laden (California Housing Dataset)"
    ]
   },
   {
@@ -799,6 +840,49 @@
     "model.fit(x_train_scaled, y_train)"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "59a43b36-584b-44a1-8886-794e6dae7387",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "array([ 0.85438303,  0.12254624, -0.29441013,  0.33925949, -0.00230772,\n",
+       "       -0.0408291 , -0.89692888, -0.86984178])"
+      ]
+     },
+     "execution_count": 7,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "model.coef_"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "f1d07fc0-8600-44be-b5b4-50a6352b7846",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "2.0719469373788777"
+      ]
+     },
+     "execution_count": 8,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "model.intercept_"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "37a770e9-051c-4842-b54d-75c67a5df7a2",
@@ -809,7 +893,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 7,
+   "execution_count": 9,
    "id": "84deab4f-653c-496e-b72e-76a3cedb0575",
    "metadata": {},
    "outputs": [],
@@ -827,7 +911,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 8,
+   "execution_count": 10,
    "id": "1a1e7040-cc53-4f9f-afa1-0c8c9ff1e6dd",
    "metadata": {},
    "outputs": [
@@ -835,8 +919,8 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Mittlerer quadratischer Fehler (MSE): 0.5558915986952444\n",
-      "Bestimmtheitsmaß (R²): 0.5757877060324508\n"
+      "Mittlerer quadratischer Fehler (MSE): 0.5558915986952442\n",
+      "Bestimmtheitsmaß (R²): 0.575787706032451\n"
      ]
     }
    ],
@@ -857,13 +941,13 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 9,
+   "execution_count": 11,
    "id": "16270434-db30-472f-ab59-033142994c49",
    "metadata": {},
    "outputs": [
     {
      "data": {
-      "image/png": "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",
+      "image/png": "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",
       "text/plain": [
        "<Figure size 640x480 with 1 Axes>"
       ]
@@ -879,6 +963,14 @@
     "plt.title(\"Tatsächliche vs. Vorhergesagte Immobilienpreise\")\n",
     "plt.show()"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "ac7dd956-5204-46e0-b6ee-8fc767ed4f30",
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {
@@ -897,7 +989,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.12.9"
+   "version": "3.9.6"
   },
   "widgets": {
    "application/vnd.jupyter.widget-state+json": {
diff --git a/docs/1intro/evaluation.rst b/docs/1intro/evaluation.rst
index 231db0b..75542f2 100644
--- a/docs/1intro/evaluation.rst
+++ b/docs/1intro/evaluation.rst
@@ -101,7 +101,7 @@ Accuracy (Genauigkeit)
 Prozentsatz der korrekten Vorhersagen:
 
 .. math::
-   \text{Accuracy} = \frac{\text{Anzahl der korrekten Vorhersagen}}{\text{Gesamtanzahl der Vorhersagen}}
+   \text{Accuracy} = \frac{\text{Anzahl der korrekten Vorhersagen}}{\text{Gesamtanzahl der Vorhersagen}} = \frac{\text{TP}+\text{TN}}{all}
 
 .. warning:
    Accuracy hat eine Einschränkung bei (stark) unausgewogenen Datensätzen, da
diff --git a/docs/1intro/index.rst b/docs/1intro/index.rst
index bc3fd0d..8c6bb43 100644
--- a/docs/1intro/index.rst
+++ b/docs/1intro/index.rst
@@ -46,4 +46,5 @@ Gliederung
    linear-regression
    case-study
    reflection
+   usecase-pitfalls
    extra
diff --git a/docs/1intro/linear-regression.rst b/docs/1intro/linear-regression.rst
index 489da15..d37c3ba 100644
--- a/docs/1intro/linear-regression.rst
+++ b/docs/1intro/linear-regression.rst
@@ -71,4 +71,4 @@ Schritte zur Implementierung eines ML-Modells
 4. **Modell evaluieren und Interpretation**
 
    * Bewertung der Modellperformance auf dem Testdatensatz.
-   * Interpretation der wichtigsten Einflussgrößen.
+   * Interpretation der wichtigsten Einflussgrößen.
\ No newline at end of file
diff --git a/docs/2deep_ml_ops/usecase-pitfalls.rst b/docs/1intro/usecase-pitfalls.rst
similarity index 77%
rename from docs/2deep_ml_ops/usecase-pitfalls.rst
rename to docs/1intro/usecase-pitfalls.rst
index e413749..19bdde9 100644
--- a/docs/2deep_ml_ops/usecase-pitfalls.rst
+++ b/docs/1intro/usecase-pitfalls.rst
@@ -11,6 +11,9 @@ Use Case Fallgruben – Diskussion zur Kundenabwanderung (Customer Churn Predict
 
 Ein Telekommunikationsunternehmen möchte vorhersagen, ob ein Kunde abwandern wird oder nicht. Die Lernenden sollen erarbeiten:
 
+Relevante Gedankengänge: 
+~~~~~~~~~~~~~~~~~~~~~~~~~
+
 1. **Wie definiert man Kundenabwanderung?**
 
 - z.B. Vertragskunde: kündigt in den nächsten drei Monaten seinen Vertrag
@@ -18,13 +21,15 @@ Ein Telekommunikationsunternehmen möchte vorhersagen, ob ein Kunde abwandern wi
 
 2. **Welche Daten sind relevant?**
 
-2. **Wie sollten Features gestaltet werden?** (z. B. Anrufhäufigkeit, Vertragslaufzeit, Reklamationen?)
+3. **Wie sollten Features gestaltet werden?** 
+
+Tip: Anrufhäufigkeit, Vertragslaufzeit, Reklamationen?
 
-3. **Welche ML-Methoden eignen sich?** (Klassifikation, Feature Engineering)
+4. **Welche ML-Methoden eignen sich?** (Klassifikation, Feature Engineering)
 
-4. **Wie validiert man das Modell?**
+5. **Wie validiert man das Modell?**
 
-5. **Welche Herausforderungen können auftreten?**
+6. **Welche Herausforderungen können auftreten?**
 
 **Diskussionsfragen:**
 
diff --git a/docs/2deep_ml_ops/cnn-beispiel.ipynb b/docs/2deep_ml_ops/cnn-beispiel.ipynb
new file mode 100644
index 0000000..6792a51
--- /dev/null
+++ b/docs/2deep_ml_ops/cnn-beispiel.ipynb
@@ -0,0 +1,1066 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "53b766a3-8e54-40b0-92ba-12fcbe9025ec",
+   "metadata": {},
+   "source": [
+    "# Praxisbeispiel - Bildklassifikation mit CNNs\n",
+    "\n",
+    "## Ziel:\n",
+    "Einführung in Convolutional Neural Networks mit TensorFlow/Keras anhand eines Bildklassifikationsproblems.\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "d4dee2c8-8af5-4492-ae60-87e3181697b6",
+   "metadata": {},
+   "source": [
+    "## 1. Bibliotheken laden"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "bcb16094-dffa-4960-be59-37041ef1cd94",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import tensorflow as tf\n",
+    "from tensorflow.keras import layers, models\n",
+    "from tensorflow.keras.datasets import mnist\n",
+    "import matplotlib.pyplot as plt"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "d0f3b447-e147-4440-9d57-510303f0b327",
+   "metadata": {},
+   "source": [
+    "## 2. Daten laden"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "722a1383-267a-4526-8832-721336a47f04",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
+      "\u001b[1m11490434/11490434\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 0us/step \n"
+     ]
+    }
+   ],
+   "source": [
+    "(x_train, y_train), (x_test, y_test) = mnist.load_data()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "468189a4-a053-418e-b880-9ae50e5b5de0",
+   "metadata": {},
+   "source": [
+    "## 3. Daten einsehen"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "ccf8205f-d371-455b-8d92-e2ebf407f635",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "array([[[0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        ...,\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0]],\n",
+       "\n",
+       "       [[0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        ...,\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0]],\n",
+       "\n",
+       "       [[0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        ...,\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0]],\n",
+       "\n",
+       "       ...,\n",
+       "\n",
+       "       [[0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        ...,\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0]],\n",
+       "\n",
+       "       [[0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        ...,\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0]],\n",
+       "\n",
+       "       [[0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        ...,\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0],\n",
+       "        [0, 0, 0, ..., 0, 0, 0]]], dtype=uint8)"
+      ]
+     },
+     "execution_count": 4,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "x_train"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "d990992b-150d-4473-af09-0a366ef7cff0",
+   "metadata": {},
+   "source": [
+    "## 4. Daten vorbereiten (Data Preprocessing)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "620c9750-9f53-4071-be66-05a251ab8a4d",
+   "metadata": {},
+   "source": [
+    "### Normalisierung\n",
+    "\n",
+    "Normalisieren von Bilderdaten ist wiederum anders als von numerischen Daten. \n",
+    "\n",
+    "Der Grund für die Normalisierung der Bilder(daten) ist die Vermeidung der Möglichkeit von explodierenden Gradienten aufgrund des großen Pixelbereichs [0, 255] und die Verbesserung der Konvergenzgeschwindigkeit. \n",
+    "Daher kann entweder  \n",
+    "1. man jedes Bild normalisieren, so dass der Pixelbereich sich in [-1, 1] befindet \n",
+    "oder\n",
+    "2. man teilt jeden Wert durch den maximalen Pixelwert, d.h. 255, so dass der Bereich der Pixel im Bereich [0, 1] liegt.\n",
+    "\n",
+    "Ein weiterer Grund für die Normalisierung von Bilddaten ist wenn man Transfer Learning verwendet. \n",
+    "Wenn z. B. ein bereits trainiertes Modell verwendet wird, das mit Bildern trainiert wurde, deren Pixel im Bereich [0, 1] liegen, sollte man sicherstellen, dass die neuen Werte, die man dem Modell liefert, im gleichen Bereich liegen. Andernfalls werden die Ergebnisse verfälscht werden."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "6bcbf900-e870-4dbb-ab9f-1de80f2377a0",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Normalisierung\n",
+    "x_train, x_test = x_train / 255.0, x_test / 255.0  "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "07dee44a-37e2-4dd6-a6ae-9b8bcb7c4bf3",
+   "metadata": {},
+   "source": [
+    "### Dimension erweitern\n",
+    "\n",
+    "Zuden existierenden Dimensionen der Bilddaten fügen wir eine neue Dimension hinzu. \n",
+    "Diese neue Dimension stellt die Anzahl der in den Daten vorhandenen Kanäle dar.\n",
+    "\n",
+    "Bei Farbbildern wären dies 3 Kanäle, die den roten, grünen und blauen Kanal darstellen. \n",
+    "In diesem Fall handelt es sich um Schwarz-Weiß-Bilder, so dass nur 1 Kanal vorhanden ist."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "54f9b079-3d31-4b6e-bcda-54f9b525c6f0",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Dimension erweitern\n",
+    "x_train = x_train[..., tf.newaxis]  \n",
+    "x_test = x_test[..., tf.newaxis]"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "614c8498-fdb3-41ca-807c-79c395860466",
+   "metadata": {},
+   "source": [
+    "## 5. Modell definieren"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "d0c0b7e5-dd65-45a3-a53d-843b89266edb",
+   "metadata": {},
+   "source": [
+    "CNN Layers:\n",
+    "- Conv2D:\n",
+    "\n",
+    "\n",
+    "  Die am häufigsten verwendete Art der Faltung ist die 2D-Faltungsschicht und wird üblicherweise als conv2D abgekürzt. Ein Filter oder ein Kernel in einer conv2D-Schicht „gleitet“ über die 2D-Eingangsdaten und führt eine elementweise Multiplikation durch. Das Ergebnis ist die Summierung der Ergebnisse zu einem einzigen Ausgabepixel.\n",
+    "\n",
+    "  Parameter bei der Erstellung einer Conv2D Schicht:\n",
+    "\n",
+    "  1. 32: Anzahl von Filtern in dieser Convolution-Schicht. Hierfür wird immer empfohlen, Potenzen von 2 als Werte zu verwenden.\n",
+    "  2. (3, 3): bestimmt die Dimensionen des Kernels. Übliche Abmessungen sind 1×1, 3×3, 5×5 oder 7×7, entsprechend als (1, 1), (3, 3), (5, 5) oder (7, 7)-Tupel übergeben.\n",
+    "     Es muss hier eine ganze Zahl oder ein Tupel/Liste von 2 ganzen Zahlen, die die Höhe und Breite des 2D-Faltungsfensters angeben. Zudem muss dieser Parameter eine ungerade ganze Zahl sein.\n",
+    "  4. activation=\"..\": gibt den Namen der Aktivierungsfunktion an, die nach der Faltung/convolution verwendet werden soll. (siehe unten)\n",
+    "    \n",
+    "  \n",
+    "- MaxPooling2D (more details: https://www.geeksforgeeks.org/cnn-introduction-to-pooling-layer/?ref=header_outind) \n",
+    "\n",
+    "  Die Pooling-Schicht wird in CNNs verwendet, um die räumlichen Dimensionen (Breite und Höhe) der eingegebenen Merkmalskarten zu reduzieren und gleichzeitig die wichtigsten Informationen beizubehalten. Dabei wird ein zweidimensionaler Filter über jeden Kanal einer Merkmalskarte gezogen und die Merkmale innerhalb des vom Filter abgedeckten Bereichs zusammengefasst.\n",
+    "\n",
+    "  Zudem hilft es die Dimensionalität zu verringern, da Pooling-Schichten die räumliche Größe der Feature-Matrix reduzieren, somit die Anzahl der Parameter und Berechnungen im Network verringert wird. So wird das Modell schneller und effizienter. Außerdem trägt die Reduzierung der räumlichen Dimensionen dazu bei, Overfitting zu verhindern. \n",
+    "\n",
+    "- Flatten:\n",
+    "\n",
+    "\n",
+    "  Eine flache Schicht des neuronalen Netzes wird verwendet, um die mehrdimensionale Ausgabe der vorhergehenden Schicht in ein eindimensionales Feld umzuwandeln, bevor sie zur weiteren Verarbeitung in eine vollständig verbundene Schicht (dense layers) eingespeist wird.\n",
+    "\n",
+    "  Zudem reduziert es die Dimension in den Daten und vereinfacht die Modellarchitektur.\n",
+    "\n",
+    "  \n",
+    "- Dense:\n",
+    "  Die Dense Schicht ist eine vollständig verbundene Schicht."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "5aa5e504-e04f-4907-8677-4ce8335d992a",
+   "metadata": {},
+   "source": [
+    "### (typische) Aktivierungsfunktionen \n",
+    "- relu:\n",
+    "  \n",
+    "  Die ReLU-Aktivierungsfunktion wird verwendet, um Nichtlinearität in ein neuronales Netz einzuführen. Sie trägt dazu bei, das Problem des verschwindenden Gradienten beim Training von Modellen des maschinellen Lernens zu entschärfen, und ermöglicht es neuronalen Netzen, komplexere Beziehungen in Daten zu lernen.\n",
+    "  Wenn eine Modelleingabe positiv ist, gibt die ReLU-Funktion denselben Wert aus. Wenn eine Modelleingabe negativ ist, gibt die ReLU-Funktion den Wert Null aus.\n",
+    "\n",
+    "- softmax:\n",
+    "\n",
+    "  Die Softmax-Funktion, die häufig in der letzten Schicht eines neuronalen Netzmodells für Klassifizierungsaufgaben verwendet wird, wandelt rohe Ausgabeergebnisse - auch als Logits bekannt - in Wahrscheinlichkeiten um, indem sie den Exponentialwert jeder Ausgabe nimmt und diese Werte normalisiert, indem sie durch die Summe aller Exponentialwerte dividiert wird.\n",
+    "\n",
+    "  https://botpenguin.com/glossary/softmax-function"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "169980df-2df4-42af-b2b9-eb2cf15671dc",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "/Users/minye/Documents/gitHub/ainvone/cusy-gfu/gfu/schulung-ki_von_grund_auf/01-Einfuehrung-in-KI/.venv312/lib/python3.12/site-packages/keras/src/layers/convolutional/base_conv.py:107: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.\n",
+      "  super().__init__(activity_regularizer=activity_regularizer, **kwargs)\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Modell definieren\n",
+    "model = models.Sequential(\n",
+    "   [\n",
+    "       layers.Conv2D(32, (3, 3), activation=\"relu\", input_shape=(28, 28, 1)),\n",
+    "       layers.MaxPooling2D((2, 2)),\n",
+    "       layers.Conv2D(64, (3, 3), activation=\"relu\"),\n",
+    "       layers.MaxPooling2D((2, 2)),\n",
+    "       layers.Conv2D(64, (3, 3), activation=\"relu\"),\n",
+    "       layers.Flatten(),\n",
+    "       layers.Dense(64, activation=\"relu\"),\n",
+    "       layers.Dense(10, activation=\"softmax\"),\n",
+    "   ]\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "0dcef3d2-9b6e-4c35-b0af-24816e1fefac",
+   "metadata": {},
+   "source": [
+    "Gute Visualisierungen solch ähnlicher Struktur können hier gefunden werden:\n",
+    "\n",
+    "https://miro.medium.com/v2/resize:fit:1400/format:webp/1*vkQ0hXDaQv57sALXAJquxA.jpeg\n",
+    "\n",
+    "https://miro.medium.com/v2/resize:fit:1400/format:webp/1*uAeANQIOQPqWZnnuH-VEyw.jpeg\n",
+    "\n",
+    "(Credits an Sumit Saha: https://medium.com/towards-data-science/a-comprehensive-guide-to-convolutional-neural-networks-the-eli5-way-3bd2b1164a53)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "cc005536-b5d2-43a4-a27a-195b9fc0a6ff",
+   "metadata": {},
+   "source": [
+    "## 6. Modell kompilieren"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "0af77c33-c547-47b5-8549-f77169a4585d",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Modell kompilieren\n",
+    "model.compile(\n",
+    "   optimizer=\"adam\",\n",
+    "   loss=\"sparse_categorical_crossentropy\",\n",
+    "   metrics=[\"accuracy\"],\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "322f7757-10d8-48b6-8b6c-89bdd6f83d3f",
+   "metadata": {},
+   "source": [
+    "## 7. Training"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "159676e6-736b-473b-93a0-1cc4e759a9f6",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch 1/5\n",
+      "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 7ms/step - accuracy: 0.5201 - loss: 1.3256 - val_accuracy: 0.9181 - val_loss: 0.2701\n",
+      "Epoch 2/5\n",
+      "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 6ms/step - accuracy: 0.9247 - loss: 0.2469 - val_accuracy: 0.9536 - val_loss: 0.1489\n",
+      "Epoch 3/5\n",
+      "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 6ms/step - accuracy: 0.9534 - loss: 0.1544 - val_accuracy: 0.9650 - val_loss: 0.1072\n",
+      "Epoch 4/5\n",
+      "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 7ms/step - accuracy: 0.9662 - loss: 0.1099 - val_accuracy: 0.9733 - val_loss: 0.0798\n",
+      "Epoch 5/5\n",
+      "\u001b[1m1875/1875\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 6ms/step - accuracy: 0.9714 - loss: 0.0891 - val_accuracy: 0.9786 - val_loss: 0.0680\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "<keras.src.callbacks.history.History at 0x16a668200>"
+      ]
+     },
+     "execution_count": 9,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "model.fit(x_train, y_train, epochs=5, validation_data=(x_test, y_test))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "54e16a32-c8c8-46c5-89a8-d9ed94151f50",
+   "metadata": {
+    "collapsed": true,
+    "jupyter": {
+     "outputs_hidden": true
+    }
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "[array([[[[-1.22406550e-01, -1.04940124e-01,  1.05540431e+00,\n",
+       "            7.33939633e-02, -1.00426063e-01, -5.34634618e-03,\n",
+       "            7.14282170e-02, -1.25062847e+00,  6.30540550e-02,\n",
+       "           -2.37185694e-02, -7.45562315e-02, -1.64713878e-02,\n",
+       "           -1.12705171e+00, -1.32961959e-01, -5.87262250e-02,\n",
+       "            1.01219498e-01,  6.92881346e-01,  1.01272047e+00,\n",
+       "            1.19992472e-01,  1.23000138e-01, -3.42749394e-02,\n",
+       "           -9.46973711e-02,  4.86068949e-02,  1.68845132e-02,\n",
+       "           -1.17901556e-01, -4.75684665e-02, -1.04668069e+00,\n",
+       "           -9.32694674e-02, -9.69426095e-01,  9.41504955e-01,\n",
+       "            1.00876343e+00,  1.07508950e-01]],\n",
+       " \n",
+       "         [[ 6.25128299e-02,  1.03033878e-01,  1.63415444e+00,\n",
+       "           -5.86562604e-02,  6.17862120e-03,  6.60237372e-02,\n",
+       "           -8.15676749e-02, -1.47776830e+00, -1.31476879e-01,\n",
+       "           -1.34139717e-01,  1.89567413e-02,  1.34701788e-01,\n",
+       "           -1.18035138e+00, -3.27071249e-02,  5.31418025e-02,\n",
+       "           -9.48082097e-03,  1.22419310e+00,  1.30111873e+00,\n",
+       "            9.74448249e-02,  9.95283425e-02,  1.48501322e-02,\n",
+       "            1.00815147e-01,  2.78599281e-02,  1.32996678e-01,\n",
+       "           -4.16899063e-02,  9.56013501e-02, -1.41377544e+00,\n",
+       "            5.99457696e-02, -1.19312155e+00,  1.50736153e+00,\n",
+       "            1.61045134e+00, -9.09326226e-02]],\n",
+       " \n",
+       "         [[-7.83583298e-02, -8.82351398e-02,  1.94420063e+00,\n",
+       "           -6.19674921e-02,  1.21926457e-01, -1.64310466e-02,\n",
+       "           -6.62469305e-03, -1.89023638e+00,  6.79314369e-03,\n",
+       "           -5.72794601e-02, -1.79723538e-02, -1.20034032e-01,\n",
+       "           -1.47505045e+00, -6.51922971e-02, -1.18146993e-01,\n",
+       "            1.31704301e-01,  1.69707072e+00,  1.73330379e+00,\n",
+       "           -9.68419313e-02, -6.24438561e-02,  5.91679960e-02,\n",
+       "            3.02885417e-02, -1.17697634e-01,  4.59196344e-02,\n",
+       "            8.85993317e-02, -1.74901017e-03, -1.84632277e+00,\n",
+       "           -1.35182694e-01, -1.48471940e+00,  1.83147061e+00,\n",
+       "            1.73361361e+00, -5.30065596e-02]]],\n",
+       " \n",
+       " \n",
+       "        [[[-9.57763493e-02,  1.06843095e-02,  1.75813472e+00,\n",
+       "           -1.29688382e-01, -7.96640888e-02,  9.84542742e-02,\n",
+       "            1.28048778e-01, -1.80853605e+00, -1.17492847e-01,\n",
+       "           -3.03778518e-02,  4.00241576e-02, -8.16433206e-02,\n",
+       "           -1.90246665e+00, -6.18674979e-02,  1.03265770e-01,\n",
+       "           -9.87912118e-02,  1.41587746e+00,  1.59109902e+00,\n",
+       "           -4.26441692e-02, -1.09283321e-01,  1.34045720e-01,\n",
+       "            6.59470335e-02, -9.85139515e-03, -5.01345322e-02,\n",
+       "           -2.21411921e-02,  1.61392754e-03, -1.82987559e+00,\n",
+       "           -2.80206427e-02, -1.82214963e+00,  1.68708861e+00,\n",
+       "            1.71702063e+00, -1.81070566e-02]],\n",
+       " \n",
+       "         [[-1.21809751e-01,  3.73796299e-02,  2.49048185e+00,\n",
+       "           -8.52861255e-02,  9.44486335e-02,  1.21906884e-01,\n",
+       "           -1.11697055e-02, -1.82091975e+00,  5.71683533e-02,\n",
+       "            8.27445015e-02, -6.15169182e-02,  6.67865109e-03,\n",
+       "           -1.97300541e+00,  1.92017406e-02,  9.39941406e-03,\n",
+       "           -5.58954440e-02,  2.34084105e+00,  2.49039578e+00,\n",
+       "            1.05479367e-01,  1.11193741e-02, -5.72169982e-02,\n",
+       "           -9.02368352e-02, -8.29203147e-03,  4.14377116e-02,\n",
+       "            2.34825872e-02, -6.77365437e-02, -1.78600693e+00,\n",
+       "            5.83842322e-02, -1.64335132e+00,  2.46693969e+00,\n",
+       "            2.24549413e+00, -7.59055838e-02]],\n",
+       " \n",
+       "         [[-2.31350586e-02,  4.71984670e-02,  1.97533929e+00,\n",
+       "            6.71978071e-02, -7.17302039e-02, -2.24373341e-02,\n",
+       "            5.07648801e-03, -2.12082839e+00,  8.82488862e-02,\n",
+       "           -7.06230700e-02, -1.48546234e-01,  1.33056059e-01,\n",
+       "           -2.07195067e+00, -1.08200125e-01, -3.93704139e-02,\n",
+       "           -5.73396087e-02,  2.17356730e+00,  2.17443180e+00,\n",
+       "            5.66516481e-02,  4.58226390e-02, -5.51303923e-02,\n",
+       "            9.86828879e-02, -1.33762971e-01, -9.46188718e-02,\n",
+       "           -3.34561355e-02, -1.92110725e-02, -1.94994235e+00,\n",
+       "           -2.60656327e-02, -2.00395918e+00,  2.10578084e+00,\n",
+       "            1.85894048e+00,  3.11964843e-02]]],\n",
+       " \n",
+       " \n",
+       "        [[[ 6.48115575e-02,  4.62496839e-02,  1.31863916e+00,\n",
+       "            6.87884986e-02,  2.86218897e-02,  5.42872176e-02,\n",
+       "           -1.51097635e-02, -1.52965164e+00,  2.06916425e-02,\n",
+       "           -1.74260139e-02, -1.67195164e-02, -3.93627696e-02,\n",
+       "           -1.71877074e+00,  3.72256860e-02,  1.02261432e-01,\n",
+       "            5.50065525e-02,  1.47071218e+00,  1.35496914e+00,\n",
+       "            1.12139270e-01,  7.77536109e-02, -1.66937225e-02,\n",
+       "           -2.74120513e-02, -3.52033749e-02,  4.27518710e-02,\n",
+       "            2.96470895e-02,  7.74797425e-02, -1.63591981e+00,\n",
+       "           -6.24467283e-02, -1.49033296e+00,  1.51481032e+00,\n",
+       "            1.43474472e+00,  6.48858845e-02]],\n",
+       " \n",
+       "         [[ 1.25385031e-01,  1.93701517e-02,  1.35700619e+00,\n",
+       "           -3.29198055e-02, -1.35597497e-01,  9.18423459e-02,\n",
+       "           -1.29965737e-01, -1.41640520e+00, -9.96136293e-02,\n",
+       "           -3.75120044e-02, -3.45025845e-02, -1.12801664e-01,\n",
+       "           -1.65582597e+00,  9.59943309e-02, -3.54000740e-02,\n",
+       "           -4.21205461e-02,  1.63431835e+00,  1.57297838e+00,\n",
+       "            4.73094434e-02, -1.13333479e-01,  1.17508225e-01,\n",
+       "           -1.23712495e-01,  8.00190940e-02, -8.36999044e-02,\n",
+       "            1.23406455e-01,  1.18419128e-02, -1.26421440e+00,\n",
+       "           -5.20271473e-02, -1.36224902e+00,  1.35520363e+00,\n",
+       "            1.39786804e+00,  1.11324854e-01]],\n",
+       " \n",
+       "         [[ 1.20124757e-01,  5.89667214e-03,  9.07224596e-01,\n",
+       "           -9.21126157e-02, -1.34972453e-01,  1.06169097e-01,\n",
+       "           -5.43355346e-02, -7.48154938e-01, -8.53973702e-02,\n",
+       "            3.73354144e-02,  1.08885683e-01, -1.39362082e-01,\n",
+       "           -1.32957542e+00, -1.14915244e-01, -7.53873289e-02,\n",
+       "            1.49036013e-02,  1.12235367e+00,  1.08785880e+00,\n",
+       "           -4.18222100e-02,  4.17723209e-02, -3.86383181e-05,\n",
+       "            9.45675746e-03,  3.36362980e-02, -4.63068299e-02,\n",
+       "           -2.00620797e-02, -1.23409078e-01, -8.52762043e-01,\n",
+       "            4.43898439e-02, -1.08875692e+00,  1.16234934e+00,\n",
+       "            7.55028188e-01,  1.05900802e-01]]]], dtype=float32),\n",
+       " array([-0.00466503, -0.00600477,  0.00457475, -0.00599602, -0.00599961,\n",
+       "        -0.00517343, -0.00599918,  0.05253192, -0.00600454, -0.00600148,\n",
+       "        -0.00359347, -0.00220906,  0.05363466, -0.00078721, -0.00600311,\n",
+       "        -0.00303029,  0.00311596,  0.00346661, -0.00630751, -0.00222704,\n",
+       "        -0.00600247, -0.00614355, -0.0059998 , -0.00600276, -0.00600332,\n",
+       "        -0.00600424,  0.04722345, -0.00599897,  0.0475119 ,  0.00465371,\n",
+       "         0.00539291, -0.0056139 ], dtype=float32),\n",
+       " array([[[[-4.09513079e-02,  6.53369576e-02,  5.56352772e-02, ...,\n",
+       "            1.78319160e-02,  3.39432918e-02, -3.58552150e-02],\n",
+       "          [ 2.47284397e-02,  3.53016071e-02,  6.65583462e-02, ...,\n",
+       "           -2.61687115e-02,  6.35163262e-02, -5.15351743e-02],\n",
+       "          [-2.22452357e-01, -4.54687215e-02, -4.80623581e-02, ...,\n",
+       "           -4.98063723e-03, -6.70940503e-02, -5.90361245e-02],\n",
+       "          ...,\n",
+       "          [-2.66597062e-01, -3.98694575e-02, -1.12784944e-01, ...,\n",
+       "           -8.07477832e-02,  7.76869878e-02, -2.90160239e-01],\n",
+       "          [-2.23186836e-01, -6.77605346e-02, -1.66173652e-02, ...,\n",
+       "           -3.00479103e-02,  5.50204478e-02, -1.05014518e-01],\n",
+       "          [-7.85470307e-02, -3.20744477e-02,  3.95169184e-02, ...,\n",
+       "           -6.43990338e-02, -8.35013518e-04,  6.09391630e-02]],\n",
+       " \n",
+       "         [[ 4.09062728e-02, -7.40245059e-02,  2.20822245e-02, ...,\n",
+       "           -2.43453868e-03,  7.34328479e-02, -6.81948885e-02],\n",
+       "          [-1.07655879e-02,  5.58332168e-02, -1.34183485e-02, ...,\n",
+       "           -1.93944741e-02, -1.23288482e-02, -3.74378860e-02],\n",
+       "          [ 2.96264142e-01, -9.46638919e-03, -5.45962274e-01, ...,\n",
+       "           -6.95578456e-02, -9.17147920e-02, -8.07937980e-02],\n",
+       "          ...,\n",
+       "          [ 4.38567430e-01, -5.05007021e-02, -5.19621611e-01, ...,\n",
+       "           -7.21395388e-02,  2.83865985e-02, -1.77612677e-01],\n",
+       "          [ 2.72482246e-01,  4.66637313e-02, -4.40830916e-01, ...,\n",
+       "           -2.00214460e-02, -6.14985153e-02, -1.48249999e-01],\n",
+       "          [ 6.05688915e-02,  3.36087979e-02, -7.32045323e-02, ...,\n",
+       "            5.90418689e-02, -2.53055207e-02,  3.52625586e-02]],\n",
+       " \n",
+       "         [[ 1.00557189e-02, -4.79275286e-02,  9.53041762e-03, ...,\n",
+       "           -4.77158539e-02, -7.76287913e-02, -8.01696628e-03],\n",
+       "          [ 6.95207641e-02,  3.19592208e-02, -5.21815345e-02, ...,\n",
+       "           -2.97903102e-02,  6.85276613e-02, -5.29903881e-02],\n",
+       "          [ 1.71262100e-01,  1.36805242e-02, -9.78173018e-01, ...,\n",
+       "            4.49864306e-02, -1.20615307e-02, -2.84896433e-01],\n",
+       "          ...,\n",
+       "          [ 1.71821341e-01,  2.18086075e-02, -9.61809039e-01, ...,\n",
+       "            5.50260358e-02, -9.68773663e-03, -3.86274487e-01],\n",
+       "          [ 1.60585135e-01,  4.09759246e-02, -7.95136929e-01, ...,\n",
+       "            2.87758000e-02,  1.87965482e-02, -3.41403008e-01],\n",
+       "          [ 1.96204372e-02,  3.06844991e-02, -2.38991342e-02, ...,\n",
+       "           -5.21317907e-02,  2.88740452e-02,  5.26520275e-02]]],\n",
+       " \n",
+       " \n",
+       "        [[[ 5.87632172e-02, -8.22747350e-02, -3.34531330e-02, ...,\n",
+       "           -6.42238185e-02, -1.10533163e-02, -6.01161979e-02],\n",
+       "          [ 4.42620032e-02,  7.78351128e-02,  6.86004013e-02, ...,\n",
+       "           -4.92207194e-03,  4.17802390e-03, -5.29544502e-02],\n",
+       "          [ 1.80385873e-01, -6.30437732e-02, -1.26061618e-01, ...,\n",
+       "            6.30052313e-02,  4.30273972e-02, -3.90959948e-01],\n",
+       "          ...,\n",
+       "          [ 2.31140316e-01, -1.80431828e-02, -2.00009853e-01, ...,\n",
+       "           -4.68120119e-03, -4.29315753e-02, -4.45163518e-01],\n",
+       "          [ 2.65494496e-01,  4.85510565e-02, -1.15276620e-01, ...,\n",
+       "            1.69637648e-03,  3.52750495e-02, -3.95026982e-01],\n",
+       "          [ 8.41122717e-02, -4.40930463e-02, -1.07017783e-02, ...,\n",
+       "           -3.00858682e-03, -3.21241841e-02,  2.35249344e-02]],\n",
+       " \n",
+       "         [[ 4.23739143e-02, -3.92357372e-02, -1.63357295e-02, ...,\n",
+       "            9.08324029e-03,  7.45074078e-02, -2.95386836e-02],\n",
+       "          [ 8.75813812e-02, -7.73116052e-02, -2.93690488e-02, ...,\n",
+       "            2.73313113e-02, -3.16706188e-02,  4.83886674e-02],\n",
+       "          [ 5.65019667e-01, -3.46408747e-02, -6.41785145e-01, ...,\n",
+       "           -4.36245799e-02, -1.66090264e-03, -4.39111143e-01],\n",
+       "          ...,\n",
+       "          [ 5.93827307e-01, -3.87401599e-03, -6.45603538e-01, ...,\n",
+       "           -2.82105263e-02,  3.12332362e-02, -4.72609550e-01],\n",
+       "          [ 4.74565119e-01, -7.93136731e-02, -6.07311606e-01, ...,\n",
+       "            2.62665581e-02,  6.36101142e-02, -5.15639246e-01],\n",
+       "          [ 6.92659616e-02, -2.31724046e-02,  4.59018424e-02, ...,\n",
+       "           -5.38128801e-02, -3.86629738e-02,  2.16949563e-02]],\n",
+       " \n",
+       "         [[-5.09778857e-02, -2.56684553e-02, -4.65940572e-02, ...,\n",
+       "           -2.24864893e-02,  5.98053634e-02, -3.33038643e-02],\n",
+       "          [-4.89353910e-02, -5.94337359e-02, -3.05621680e-02, ...,\n",
+       "           -1.74723398e-02, -1.26119517e-02,  7.22276121e-02],\n",
+       "          [ 4.27562505e-01, -7.00907260e-02, -1.51265748e-02, ...,\n",
+       "           -3.63055877e-02,  3.77160870e-02, -3.35449606e-01],\n",
+       "          ...,\n",
+       "          [ 5.34057260e-01,  5.37596922e-03,  5.90588860e-02, ...,\n",
+       "           -2.23858450e-02, -4.39759269e-02, -4.74299133e-01],\n",
+       "          [ 4.68883842e-01,  4.67453059e-03, -1.07363835e-02, ...,\n",
+       "           -5.03335521e-02,  6.17169142e-02, -3.97303224e-01],\n",
+       "          [-6.89325705e-02,  2.09207926e-02,  6.66928813e-02, ...,\n",
+       "            1.09824250e-02,  6.55553713e-02,  7.71144405e-02]]],\n",
+       " \n",
+       " \n",
+       "        [[[-1.97533462e-02, -5.83553733e-03, -6.98258951e-02, ...,\n",
+       "           -2.01653466e-02, -1.41384220e-02,  3.08275688e-02],\n",
+       "          [ 1.42864808e-02,  6.97184727e-02,  7.59560391e-02, ...,\n",
+       "           -5.29891141e-02, -4.83914688e-02,  4.91325147e-02],\n",
+       "          [-5.54303825e-03, -1.26073090e-02, -4.71151650e-01, ...,\n",
+       "           -2.78666001e-02, -8.68165642e-02, -1.18691437e-01],\n",
+       "          ...,\n",
+       "          [-5.74064963e-02, -5.98900020e-03, -6.38170838e-01, ...,\n",
+       "           -8.69315991e-04, -3.79942767e-02, -2.97842801e-01],\n",
+       "          [-2.29144637e-02,  5.93629219e-02, -4.72507209e-01, ...,\n",
+       "            4.12018448e-02,  2.71730740e-02, -3.18835378e-01],\n",
+       "          [ 8.15796331e-02,  2.51329765e-02,  5.64804859e-03, ...,\n",
+       "            2.70471610e-02, -5.53913526e-02, -5.22312755e-03]],\n",
+       " \n",
+       "         [[-2.81268992e-02, -1.71664648e-03, -2.59977784e-02, ...,\n",
+       "            3.32215242e-02, -8.31729695e-02, -8.05165321e-02],\n",
+       "          [ 7.21393004e-02,  3.44517343e-02,  8.06643534e-03, ...,\n",
+       "            1.90443415e-02,  1.78395640e-02, -2.70510036e-02],\n",
+       "          [ 2.44110003e-01, -2.84078578e-03, -4.59649205e-01, ...,\n",
+       "           -3.39153633e-02,  2.68028434e-02,  3.49525571e-01],\n",
+       "          ...,\n",
+       "          [ 3.69173735e-01,  2.78887711e-02, -5.01491487e-01, ...,\n",
+       "            6.77736988e-03, -6.90143555e-02,  2.68187821e-01],\n",
+       "          [ 3.17912370e-01,  2.90262271e-02, -4.88924950e-01, ...,\n",
+       "           -6.36181459e-02, -5.71498461e-02,  1.36232108e-01],\n",
+       "          [ 4.04094532e-02, -2.83132493e-02, -4.72990127e-04, ...,\n",
+       "            5.18598892e-02,  9.73773468e-03, -5.75573072e-02]],\n",
+       " \n",
+       "         [[-8.04608688e-02,  3.08847427e-02, -4.04163823e-02, ...,\n",
+       "            6.85387850e-02, -7.79923797e-02, -7.31467605e-02],\n",
+       "          [ 8.47627148e-02,  4.50354367e-02, -7.16433302e-03, ...,\n",
+       "            6.19982854e-02, -6.19668588e-02,  2.09368914e-02],\n",
+       "          [ 4.49168593e-01, -2.93983165e-02,  2.44198084e-01, ...,\n",
+       "           -7.84730166e-03, -2.02843752e-02,  3.40622038e-01],\n",
+       "          ...,\n",
+       "          [ 5.60770750e-01,  5.69978617e-02,  2.70948112e-01, ...,\n",
+       "           -5.29176369e-02,  5.11268228e-02,  2.42241070e-01],\n",
+       "          [ 4.79552686e-01, -6.09705336e-02,  2.10609511e-01, ...,\n",
+       "            7.63696060e-02, -1.84087232e-02,  1.77408844e-01],\n",
+       "          [-3.74305770e-02,  4.71676774e-02, -5.75387478e-02, ...,\n",
+       "            3.90911661e-02, -2.25753058e-02,  3.31210904e-02]]]],\n",
+       "       dtype=float32),\n",
+       " array([ 5.0731726e-02, -6.5370905e-03,  7.3645994e-02, -3.1689333e-03,\n",
+       "        -3.0370019e-02,  5.1700506e-02, -4.8823613e-03, -1.6101491e-02,\n",
+       "        -7.9340730e-03,  4.9064219e-02, -7.0074289e-03, -6.0041053e-03,\n",
+       "        -6.0031614e-03,  4.9787950e-02, -2.4142025e-02, -2.1615267e-02,\n",
+       "        -6.0047242e-03, -7.7282340e-05, -1.0096843e-02, -6.0587190e-03,\n",
+       "         8.4531337e-02,  2.0987438e-02, -3.1731434e-02, -4.8332978e-03,\n",
+       "         3.9843556e-02, -3.6939710e-02, -5.9687099e-03, -6.0045891e-03,\n",
+       "        -6.0048811e-03, -7.1099498e-03, -6.1658891e-03, -1.1944458e-02,\n",
+       "        -2.7621938e-02, -2.5463229e-02, -6.3406507e-04, -2.6880821e-02,\n",
+       "        -7.4103735e-03, -4.9300166e-03, -1.1226194e-02, -3.5644656e-03,\n",
+       "        -1.4320927e-03,  4.3262521e-04, -1.4444536e-02, -1.0344965e-02,\n",
+       "         6.4538099e-02, -8.7140584e-03, -1.5734933e-02,  8.1914388e-02,\n",
+       "        -4.0456492e-02,  6.5072395e-02, -1.2234733e-02,  2.9029636e-02,\n",
+       "        -6.0050460e-03, -3.3119125e-03, -6.0045579e-03, -2.1562304e-02,\n",
+       "        -3.5139427e-03, -5.8292379e-03, -6.0048150e-03, -1.1523294e-03,\n",
+       "        -1.0892602e-02, -6.0025603e-03, -3.3935241e-03,  2.6472285e-02],\n",
+       "       dtype=float32),\n",
+       " array([[[[-6.77588284e-02,  2.20274855e-03,  2.99741030e-02, ...,\n",
+       "           -1.02044838e-02, -1.58815347e-02,  2.59737298e-02],\n",
+       "          [-4.84939553e-02, -2.06980687e-02, -5.60233518e-02, ...,\n",
+       "           -6.21065237e-02, -8.72860942e-03,  3.03782523e-02],\n",
+       "          [ 6.16623163e-02,  3.21431011e-02,  3.25699784e-02, ...,\n",
+       "            1.17047809e-01, -2.47664955e-02, -4.19078358e-02],\n",
+       "          ...,\n",
+       "          [ 4.48789671e-02,  2.42642276e-02, -1.59621984e-03, ...,\n",
+       "           -1.86351314e-02,  6.38029352e-02, -3.06290830e-03],\n",
+       "          [-2.87983976e-02, -7.94108957e-03,  6.51264787e-02, ...,\n",
+       "           -2.74584629e-02, -3.73801589e-02,  1.90621875e-02],\n",
+       "          [-2.16469709e-02,  4.30435315e-02,  5.76777421e-02, ...,\n",
+       "           -3.32496017e-02, -8.54106173e-02, -3.01648788e-02]],\n",
+       " \n",
+       "         [[-4.78787906e-02,  3.23773287e-02, -6.11309370e-04, ...,\n",
+       "            1.26614094e-01,  2.36432087e-02, -5.90588190e-02],\n",
+       "          [ 3.59422364e-03,  5.86653166e-02,  3.12487260e-02, ...,\n",
+       "            6.69940934e-02, -5.80463260e-02, -6.56544045e-02],\n",
+       "          [ 1.18795559e-02, -9.21184495e-02, -5.56785092e-02, ...,\n",
+       "           -2.49458954e-01,  2.80045364e-02,  1.99826416e-02],\n",
+       "          ...,\n",
+       "          [-6.72316328e-02, -5.85822202e-02,  3.74447480e-02, ...,\n",
+       "           -4.17119935e-02,  9.38113488e-04, -2.45996173e-02],\n",
+       "          [-5.52432574e-02, -6.02680854e-02, -6.05881698e-02, ...,\n",
+       "           -2.15650704e-02,  2.96953805e-02, -3.42316516e-02],\n",
+       "          [ 6.47027269e-02, -2.91282497e-02, -6.04169331e-02, ...,\n",
+       "           -2.05045994e-02,  5.47368824e-02,  4.88121063e-02]],\n",
+       " \n",
+       "         [[-1.13176079e-02, -7.21214488e-02, -1.98519975e-02, ...,\n",
+       "            1.88939333e-01, -7.92412311e-02, -6.13443330e-02],\n",
+       "          [ 6.38257861e-02,  1.56186130e-02, -5.36187924e-02, ...,\n",
+       "           -4.67634201e-02,  6.12646043e-02, -4.85889390e-02],\n",
+       "          [-2.96882465e-02,  1.51898526e-02, -9.12287235e-02, ...,\n",
+       "           -1.80930674e-01,  4.60227691e-02, -4.13853191e-02],\n",
+       "          ...,\n",
+       "          [ 1.03720725e-02,  3.70528065e-02,  3.20532173e-03, ...,\n",
+       "            4.96021798e-03, -3.74522507e-02, -5.94452880e-02],\n",
+       "          [ 5.67702465e-02, -5.15246540e-02, -3.40847229e-03, ...,\n",
+       "           -1.33436536e-02, -2.18520053e-02,  8.93314183e-03],\n",
+       "          [ 2.32600607e-02, -3.02573126e-02, -2.01285742e-02, ...,\n",
+       "           -2.17192560e-01, -3.92585881e-02, -1.96688734e-02]]],\n",
+       " \n",
+       " \n",
+       "        [[[-1.31473243e-02,  4.30563604e-03, -8.74016955e-02, ...,\n",
+       "            2.50245519e-02,  4.73613776e-02, -7.22233206e-02],\n",
+       "          [-3.19643766e-02, -4.31475230e-02,  5.31325787e-02, ...,\n",
+       "            4.96112229e-03, -6.35493025e-02, -1.41577478e-02],\n",
+       "          [-4.46854644e-02, -6.16688319e-02, -1.08877558e-03, ...,\n",
+       "           -1.07957714e-03,  8.01343005e-03, -2.04664450e-02],\n",
+       "          ...,\n",
+       "          [ 5.05997390e-02,  3.32729816e-02,  1.33315176e-02, ...,\n",
+       "           -2.05177888e-02, -4.95410338e-02, -2.29872596e-02],\n",
+       "          [-4.66450453e-02,  5.74672632e-02,  3.70416567e-02, ...,\n",
+       "            4.40301299e-02,  5.70785031e-02, -1.92973036e-02],\n",
+       "          [ 9.16888565e-03,  4.06818874e-02, -7.77892247e-02, ...,\n",
+       "            2.50012696e-01,  1.06723420e-02,  3.93631123e-02]],\n",
+       " \n",
+       "         [[-1.85088329e-02, -4.42485772e-02,  7.18603516e-03, ...,\n",
+       "           -1.07027076e-01, -7.43248090e-02,  3.62998582e-02],\n",
+       "          [ 5.17364554e-02, -1.96498986e-02, -6.19538538e-02, ...,\n",
+       "            4.23023626e-02,  1.13501688e-02, -6.03644103e-02],\n",
+       "          [ 5.76595366e-02, -7.50290230e-02, -4.94544692e-02, ...,\n",
+       "           -2.61475384e-01, -5.48628755e-02, -2.71031186e-02],\n",
+       "          ...,\n",
+       "          [-3.99222374e-02,  5.60345873e-02, -3.66110802e-02, ...,\n",
+       "           -5.38463555e-02, -4.46803030e-03,  5.09350859e-02],\n",
+       "          [ 5.64508401e-02,  5.68455681e-02,  4.77742329e-02, ...,\n",
+       "           -4.49834764e-02,  1.34397869e-03,  8.74625053e-03],\n",
+       "          [ 4.74369787e-02,  3.86862457e-02,  8.30656476e-03, ...,\n",
+       "            1.11709051e-01,  5.02744596e-03,  1.54393408e-02]],\n",
+       " \n",
+       "         [[-5.12286834e-02, -4.57517877e-02,  3.29876170e-02, ...,\n",
+       "            9.96750444e-02, -5.28524891e-02,  4.56503369e-02],\n",
+       "          [ 3.89702581e-02, -1.06641045e-02, -3.81669998e-02, ...,\n",
+       "           -1.46490242e-02,  2.37572175e-02, -3.21111120e-02],\n",
+       "          [-4.46778275e-02, -4.90375273e-02, -8.67610574e-02, ...,\n",
+       "            4.07065749e-01, -7.27922842e-02,  4.27938364e-02],\n",
+       "          ...,\n",
+       "          [-4.30758521e-02, -5.89693263e-02, -3.76514606e-02, ...,\n",
+       "           -9.67393722e-03,  5.87980039e-02, -2.92121363e-03],\n",
+       "          [ 5.90541121e-03, -5.55726923e-02, -2.65220068e-02, ...,\n",
+       "           -3.33338417e-02,  4.62677851e-02, -1.28199011e-02],\n",
+       "          [-5.91470394e-03, -1.81701519e-02, -9.07925963e-02, ...,\n",
+       "            7.21939579e-02, -7.48036662e-03, -5.05576283e-02]]],\n",
+       " \n",
+       " \n",
+       "        [[[-3.85871977e-02, -8.68022814e-02, -2.40917522e-02, ...,\n",
+       "           -1.56738572e-02,  3.65018286e-02, -1.85995977e-02],\n",
+       "          [-3.78075913e-02, -2.95530465e-02,  2.18991935e-02, ...,\n",
+       "            2.44227685e-02,  3.81797776e-02, -3.28961276e-02],\n",
+       "          [ 3.09773982e-02, -1.24242818e-02, -3.42548452e-02, ...,\n",
+       "            9.25620645e-03, -2.92590801e-02, -4.68145870e-02],\n",
+       "          ...,\n",
+       "          [-6.85492381e-02, -1.40666896e-02,  4.94359210e-02, ...,\n",
+       "           -3.93450893e-02, -1.93639342e-02,  6.83873445e-02],\n",
+       "          [ 5.05576318e-04, -6.19990285e-03, -2.55681425e-02, ...,\n",
+       "           -5.44838831e-02,  2.39748675e-02, -6.64056391e-02],\n",
+       "          [-3.30624022e-02, -6.65481910e-02,  2.51229946e-02, ...,\n",
+       "            4.94755059e-02, -4.49179634e-02, -7.36099184e-02]],\n",
+       " \n",
+       "         [[ 4.23321202e-02,  2.97866408e-02,  2.33436786e-02, ...,\n",
+       "            8.53242278e-02, -7.07813026e-03,  4.14080173e-02],\n",
+       "          [-5.49987294e-02, -2.92220265e-02, -6.94556832e-02, ...,\n",
+       "            6.09453730e-02, -2.83748563e-02,  5.60829900e-02],\n",
+       "          [ 2.07184963e-02,  1.98485777e-02, -3.56923486e-03, ...,\n",
+       "            1.52495205e-01, -8.83794390e-04, -1.27912201e-02],\n",
+       "          ...,\n",
+       "          [ 2.53824182e-02, -3.50722931e-02, -3.73297557e-02, ...,\n",
+       "            2.72302888e-02, -6.72836206e-04, -2.32999809e-02],\n",
+       "          [-1.77186504e-02, -7.24398568e-02,  5.56261220e-04, ...,\n",
+       "            7.64290616e-02, -7.81513471e-03,  6.85167536e-02],\n",
+       "          [ 3.62182409e-03, -6.11639991e-02,  1.95676256e-02, ...,\n",
+       "            2.53234580e-02, -1.75850902e-04,  2.60092318e-02]],\n",
+       " \n",
+       "         [[-6.77168965e-02, -9.71147865e-02, -1.65844485e-02, ...,\n",
+       "           -2.36974046e-01,  2.54155006e-02, -3.81687433e-02],\n",
+       "          [ 2.41206270e-02, -6.52763098e-02,  2.85268873e-02, ...,\n",
+       "           -2.40547080e-02, -3.35692279e-02,  1.61384791e-02],\n",
+       "          [ 6.41730577e-02,  2.43203659e-02, -2.69184029e-03, ...,\n",
+       "            4.24523443e-01, -6.81265369e-02, -5.46571575e-02],\n",
+       "          ...,\n",
+       "          [-4.00961339e-02,  3.94167118e-02,  1.06965154e-02, ...,\n",
+       "           -5.94192185e-02,  4.29948010e-02,  7.72575615e-03],\n",
+       "          [ 3.30222361e-02, -6.37925789e-02, -2.47988850e-02, ...,\n",
+       "           -4.97528352e-02,  3.69183533e-03,  1.24212746e-02],\n",
+       "          [ 5.26041165e-02,  6.50643883e-03, -7.16269463e-02, ...,\n",
+       "            2.10398167e-01, -2.73971539e-02,  4.08791229e-02]]]],\n",
+       "       dtype=float32),\n",
+       " array([-0.00597647, -0.00592799, -0.00870279, -0.00598878, -0.01648423,\n",
+       "        -0.01458442, -0.01836243, -0.0636421 ,  0.02648541,  0.07465245,\n",
+       "        -0.00620836,  0.02147621,  0.00727089, -0.00635401, -0.01991015,\n",
+       "        -0.01080035, -0.00645889,  0.00049094, -0.01246897, -0.02521407,\n",
+       "        -0.00600487,  0.08201219, -0.0243096 , -0.01732669, -0.01840139,\n",
+       "        -0.00134726, -0.03898312, -0.01615051, -0.00679975, -0.00140235,\n",
+       "        -0.01687402, -0.01145719, -0.06657502, -0.00450062, -0.0568934 ,\n",
+       "        -0.0258188 , -0.00581816, -0.02063407, -0.00600462, -0.01115616,\n",
+       "        -0.00582715,  0.0226899 ,  0.01437778,  0.01981827, -0.032563  ,\n",
+       "        -0.00743619, -0.02184827, -0.05019416, -0.00805772, -0.03283894,\n",
+       "        -0.01049039, -0.00781332, -0.0382159 ,  0.06350283, -0.0065008 ,\n",
+       "        -0.00736187, -0.00345836, -0.00449441, -0.00792683, -0.01904942,\n",
+       "         0.04208557, -0.03795724, -0.00628128, -0.00974068], dtype=float32),\n",
+       " array([[-0.09185279, -0.08864406,  0.07367334, ...,  0.03200333,\n",
+       "          0.07789004,  0.04746935],\n",
+       "        [ 0.09739629,  0.03832531, -0.02122908, ...,  0.0439161 ,\n",
+       "         -0.02092455,  0.05045022],\n",
+       "        [-0.0880066 ,  0.09418431, -0.08355792, ..., -0.03458135,\n",
+       "         -0.02729838, -0.05673207],\n",
+       "        ...,\n",
+       "        [ 0.13907701, -0.0463694 , -0.01554128, ...,  0.21222137,\n",
+       "         -0.13243416, -0.08443874],\n",
+       "        [ 0.02594028, -0.08970061,  0.05106094, ...,  0.03016211,\n",
+       "         -0.06699792, -0.09419772],\n",
+       "        [-0.02880331,  0.00064806, -0.03517756, ..., -0.04236604,\n",
+       "          0.09366744,  0.02290797]], dtype=float32),\n",
+       " array([ 0.06808085, -0.00975904, -0.00285351,  0.05911933,  0.00888091,\n",
+       "        -0.0083362 , -0.08058061, -0.06973764,  0.07285105,  0.12510723,\n",
+       "        -0.02541726, -0.00497621,  0.09497405,  0.01185297,  0.02194878,\n",
+       "        -0.02191131,  0.01238914, -0.03589321,  0.04940122,  0.10837622,\n",
+       "        -0.03305043,  0.07701242, -0.08698685,  0.12424077, -0.01409515,\n",
+       "         0.01491846, -0.01576733, -0.03922584, -0.00502902, -0.03992312,\n",
+       "        -0.014053  , -0.00102963,  0.00205929, -0.0182695 ,  0.00493339,\n",
+       "        -0.00217036, -0.02756455, -0.04314969, -0.01223689, -0.02844172,\n",
+       "        -0.00869887, -0.07419655, -0.00817417,  0.00466795, -0.03130533,\n",
+       "        -0.04176845, -0.02984148, -0.00221793, -0.03603784, -0.0165845 ,\n",
+       "        -0.01594525,  0.01229428, -0.00780975, -0.01449327,  0.01651074,\n",
+       "        -0.03242023,  0.00524851,  0.11353564, -0.01411175, -0.00610878,\n",
+       "        -0.11137784,  0.05338038,  0.06014742, -0.02137435], dtype=float32),\n",
+       " array([[ 0.18559797, -0.27311072,  0.01527372,  0.1667958 , -0.6008345 ,\n",
+       "          0.3159435 , -0.5216771 ,  0.2750416 ,  0.01216201,  0.22444129],\n",
+       "        [-0.17745526,  0.11224291, -0.20360905, -0.09113023, -0.06051031,\n",
+       "          0.10407807,  0.1030496 , -0.08924961,  0.20556907,  0.02068579],\n",
+       "        [-0.27077624, -0.15237704,  0.15392381, -0.019488  , -0.21341197,\n",
+       "          0.04876609,  0.26545262,  0.01613875,  0.13257483,  0.0043325 ],\n",
+       "        [ 0.00524768, -0.08751301,  0.15689705,  0.27446026,  0.03383272,\n",
+       "         -0.06179533, -0.1814533 , -0.51472723,  0.2606688 , -0.35078168],\n",
+       "        [-0.29350206, -0.11415462,  0.4038854 ,  0.04978709,  0.16851981,\n",
+       "          0.24796341, -0.2639416 , -0.10832281, -0.33657458,  0.37864706],\n",
+       "        [-0.22046587,  0.15484355, -0.14180325,  0.0856993 , -0.21411614,\n",
+       "          0.12637348, -0.2716083 ,  0.11466607,  0.24692845,  0.20042904],\n",
+       "        [ 0.15338853,  0.35099944,  0.15256815, -0.36093542, -0.24002992,\n",
+       "          0.3184705 , -0.2406162 ,  0.4650115 , -0.3069936 ,  0.02111373],\n",
+       "        [ 0.05054694, -0.23012488,  0.12789647,  0.08831997,  0.08784705,\n",
+       "          0.05929139, -0.12433138,  0.46622413, -0.35942605, -0.54049563],\n",
+       "        [-0.18717451, -0.72133374, -0.0115177 , -0.35869747,  0.22846606,\n",
+       "          0.15737483,  0.15764931, -0.38010204,  0.22650003,  0.06280418],\n",
+       "        [ 0.26188982, -0.39597288, -0.5757446 ,  0.02510367,  0.2879426 ,\n",
+       "         -0.29269058,  0.1291492 , -0.5937857 ,  0.489574  ,  0.3951149 ],\n",
+       "        [ 0.02824754, -0.18750103, -0.24379265, -0.15277593, -0.2616862 ,\n",
+       "         -0.11316513, -0.18987413, -0.10279651,  0.00307356,  0.20360096],\n",
+       "        [ 0.0889672 , -0.27085298, -0.07818568, -0.04857931,  0.26699746,\n",
+       "          0.27253887,  0.13635758,  0.20185456, -0.21887997, -0.04505806],\n",
+       "        [ 0.24088947, -0.15809816, -0.30324033,  0.32222757, -0.4815252 ,\n",
+       "         -0.15549862, -0.12925379, -0.2341969 ,  0.17756256,  0.11413763],\n",
+       "        [-0.2911384 , -0.14499116,  0.0049213 ,  0.41103235, -0.2739251 ,\n",
+       "          0.20617214, -0.56356406,  0.29324108,  0.1796953 , -0.15911548],\n",
+       "        [-0.44207096,  0.5638581 , -0.09978738,  0.13820046, -0.09684686,\n",
+       "          0.07571828,  0.15660231, -0.4522957 ,  0.02295363, -0.04906129],\n",
+       "        [ 0.06692921, -0.1146377 , -0.20020963,  0.07373878,  0.175486  ,\n",
+       "         -0.22167481, -0.25405025, -0.24551588,  0.03334694,  0.00379215],\n",
+       "        [-0.15106998, -0.27643594,  0.2916736 ,  0.409728  , -0.32726678,\n",
+       "         -0.37141395, -0.24061966, -0.2830291 ,  0.42238238, -0.03705543],\n",
+       "        [-0.06238353,  0.22148384, -0.11515956, -0.2466735 ,  0.22509186,\n",
+       "          0.01145077,  0.01299847, -0.20515227, -0.12300055,  0.06140441],\n",
+       "        [ 0.40663242, -0.5287505 ,  0.08713097,  0.1811561 ,  0.3119356 ,\n",
+       "          0.11860475,  0.3004479 , -0.24464774, -0.2475557 ,  0.23569347],\n",
+       "        [ 0.18816838, -0.27300677, -0.40234625, -0.32845914, -0.33594438,\n",
+       "          0.40800986,  0.36867064, -0.12859762,  0.196742  ,  0.44018117],\n",
+       "        [-0.3253747 ,  0.07417136, -0.3145336 , -0.18829519,  0.12210796,\n",
+       "          0.07297806,  0.0066867 , -0.10749532, -0.16823493, -0.02764349],\n",
+       "        [ 0.12949178, -0.47820517,  0.27980348, -0.11410157, -0.2167273 ,\n",
+       "          0.2086147 ,  0.18950069, -0.46491635, -0.07266409, -0.0612276 ],\n",
+       "        [ 0.23653965,  0.15099631,  0.09443749, -0.52582467,  0.06388835,\n",
+       "         -0.6205905 ,  0.02477816,  0.63442653, -0.03840058, -0.20507059],\n",
+       "        [ 0.15331094, -0.5856383 ,  0.03286755, -0.34037223,  0.13611066,\n",
+       "         -0.5115831 , -0.04399058, -0.09334271,  0.5094095 , -0.09425292],\n",
+       "        [ 0.2540139 ,  0.20344292,  0.21460418, -0.21874039,  0.06295377,\n",
+       "          0.20402728,  0.08109889,  0.2739563 , -0.04060334,  0.18409622],\n",
+       "        [-0.1093787 , -0.02947157,  0.09986844,  0.29385972, -0.0125134 ,\n",
+       "          0.08148661, -0.78703356,  0.16302125, -0.08231801,  0.13777393],\n",
+       "        [ 0.23390974, -0.03894943,  0.00211752, -0.11775871,  0.07282112,\n",
+       "          0.139505  , -0.19621724, -0.09976989,  0.1795159 ,  0.07007407],\n",
+       "        [ 0.35990015,  0.3459826 , -0.49273464, -0.2828125 , -0.52876997,\n",
+       "         -0.08519766, -0.36835986,  0.33231884,  0.10246973, -0.00127755],\n",
+       "        [ 0.06828795, -0.031198  ,  0.04726082, -0.09958206, -0.11234741,\n",
+       "         -0.03712786,  0.01512796,  0.05343895,  0.15364365, -0.07676879],\n",
+       "        [-0.06499076, -0.06642399, -0.366879  , -0.30402142,  0.29198042,\n",
+       "          0.53114074, -0.14407964,  0.47127908, -0.15015964, -0.37670112],\n",
+       "        [-0.02499727, -0.08661019, -0.14099768, -0.12561761,  0.2028338 ,\n",
+       "          0.2530595 ,  0.12981778,  0.18186334, -0.15628599, -0.16664077],\n",
+       "        [ 0.15188472, -0.05055901, -0.24041085,  0.15565348, -0.00468388,\n",
+       "          0.24099864, -0.0921819 ,  0.14400557, -0.11632543,  0.20992793],\n",
+       "        [-0.14860031, -0.09105953, -0.10214144,  0.3028628 ,  0.26851135,\n",
+       "          0.06549001, -0.93654615,  0.10836818, -0.1256314 ,  0.42355675],\n",
+       "        [-0.04537524, -0.17462449, -0.12427247,  0.00481299, -0.06823428,\n",
+       "         -0.00809711,  0.12423966, -0.11359449, -0.14621122, -0.08394695],\n",
+       "        [ 0.30618694, -0.26674655, -0.27028254,  0.23045729, -0.72168577,\n",
+       "         -0.30981228,  0.15687896,  0.24613713, -0.286744  ,  0.29990444],\n",
+       "        [ 0.22940825, -0.1299561 ,  0.12769087, -0.25860775,  0.21869509,\n",
+       "          0.22318347, -0.00288847,  0.19313459, -0.00163152, -0.27683058],\n",
+       "        [-0.19264702, -0.17462036, -0.28016275, -0.22538798, -0.03631222,\n",
+       "         -0.13697451, -0.18460888,  0.0743093 ,  0.09512224,  0.19194269],\n",
+       "        [ 0.40017086,  0.292549  ,  0.13513093, -0.3339484 ,  0.269705  ,\n",
+       "         -0.4797384 ,  0.12899716, -0.47031328, -0.2271472 , -0.0085939 ],\n",
+       "        [-0.04702419, -0.07615919, -0.115963  , -0.16872914,  0.15664066,\n",
+       "         -0.03475128, -0.16791098,  0.21822177, -0.11241832,  0.02188025],\n",
+       "        [-0.01699849,  0.03249685, -0.1861632 ,  0.06627662,  0.11205025,\n",
+       "          0.1486169 , -0.0610617 , -0.09733338,  0.23441704,  0.21430926],\n",
+       "        [ 0.17754906,  0.14666772,  0.0739053 , -0.11227049,  0.02750205,\n",
+       "          0.23046485, -0.25042364, -0.12743914,  0.19553241,  0.18650456],\n",
+       "        [ 0.01413395, -0.12288608,  0.6965986 , -0.43952045,  0.11586455,\n",
+       "         -0.11987592, -0.06369293,  0.38486838, -0.55859375, -0.13545969],\n",
+       "        [-0.10305157, -0.04945757,  0.04087194, -0.14824198,  0.2462682 ,\n",
+       "          0.1411338 ,  0.1348989 , -0.16403393, -0.00815618,  0.24623275],\n",
+       "        [ 0.19837348, -0.02752618, -0.42066678,  0.03882539,  0.34261274,\n",
+       "          0.36462694,  0.5088082 ,  0.28455532, -0.2084445 , -0.3482247 ],\n",
+       "        [-0.3265298 ,  0.04716945, -0.3801287 , -0.20364554,  0.19342104,\n",
+       "         -0.24302767,  0.23731868,  0.10557338, -0.71381503, -0.01366054],\n",
+       "        [ 0.12520793, -0.1765534 , -0.1401379 , -0.15058628, -0.15790021,\n",
+       "          0.00437207, -0.21201445,  0.1876923 ,  0.11102927, -0.01035017],\n",
+       "        [ 0.47232   ,  0.19505255,  0.00116834,  0.03654027,  0.19161451,\n",
+       "         -0.26607552,  0.28115118, -0.40405995, -0.76557356,  0.1490531 ],\n",
+       "        [-0.07026906, -0.00780696, -0.0214348 , -0.12663366,  0.00270588,\n",
+       "         -0.11748821, -0.02560265,  0.17894788, -0.09310419,  0.18391229],\n",
+       "        [ 0.05815082,  0.03282784,  0.39610478,  0.00641445, -0.01176597,\n",
+       "         -0.4745875 , -0.04102951,  0.07604022, -0.1260373 ,  0.19123802],\n",
+       "        [ 0.11074285,  0.1294036 , -0.28424704, -0.14866841,  0.20542029,\n",
+       "          0.1165528 , -0.03898513, -0.17555453,  0.27962217,  0.26348287],\n",
+       "        [-0.12302679,  0.11007027,  0.2742532 , -0.04499654,  0.16735874,\n",
+       "         -0.02710269, -0.21448018,  0.08974988, -0.10053015,  0.17155947],\n",
+       "        [-0.29028526,  0.14155896, -0.5791296 , -0.40916735,  0.53818685,\n",
+       "         -0.1641448 ,  0.18189283,  0.3412483 ,  0.21541823,  0.20131546],\n",
+       "        [-0.01186826,  0.04993342, -0.23226942,  0.2225301 ,  0.10204543,\n",
+       "          0.1943843 ,  0.13412805,  0.16293979, -0.2006093 ,  0.05441086],\n",
+       "        [-0.26056543,  0.02323167, -0.0015296 , -0.11721803, -0.24186985,\n",
+       "         -0.0963889 ,  0.17938763, -0.00543331,  0.01509079,  0.18296184],\n",
+       "        [-0.49410722,  0.24523072,  0.22284663,  0.32980958, -0.13155726,\n",
+       "          0.05894878,  0.16636033,  0.09222839,  0.17611213, -0.58574736],\n",
+       "        [-0.09948803,  0.20319562, -0.21868108,  0.10254838, -0.2094825 ,\n",
+       "         -0.1692296 ,  0.13092874, -0.23704106,  0.02225009,  0.09926336],\n",
+       "        [-0.31647605,  0.37694058,  0.17469397, -0.12611651,  0.12901165,\n",
+       "         -0.41524988, -0.21000667,  0.21409419,  0.3235765 ,  0.01054287],\n",
+       "        [-0.2865008 , -0.59122723,  0.1963901 , -0.31103402, -0.35556117,\n",
+       "          0.18691869,  0.24213761,  0.00609558,  0.42919514,  0.09425966],\n",
+       "        [ 0.26849365, -0.21575768, -0.14220595, -0.06347691, -0.29176414,\n",
+       "         -0.19544017, -0.28446892, -0.19848925,  0.2535855 ,  0.17486499],\n",
+       "        [-0.2659144 , -0.04278589, -0.24608164, -0.24818967, -0.15233153,\n",
+       "          0.16368787, -0.09293711,  0.06584688, -0.0975661 , -0.19781224],\n",
+       "        [-0.05821171,  0.16471031,  0.41971296, -0.23256129,  0.14253083,\n",
+       "         -0.1455678 ,  0.10576846,  0.33960342, -0.8148478 , -0.06659503],\n",
+       "        [ 0.12745759,  0.3702152 , -0.27308476, -0.25207722, -0.41478533,\n",
+       "          0.4310856 ,  0.15967306, -0.4570679 , -0.05391107, -0.29273134],\n",
+       "        [ 0.294219  , -0.14976916, -0.56055194, -0.40888074,  0.1531452 ,\n",
+       "          0.3659002 , -0.32704675, -0.38213524, -0.17781752,  0.2353824 ],\n",
+       "        [-0.05576838,  0.00892646, -0.07968741, -0.1295656 , -0.09649703,\n",
+       "         -0.01505735,  0.22799006,  0.2427742 ,  0.1689535 , -0.00908666]],\n",
+       "       dtype=float32),\n",
+       " array([-0.00490798, -0.01904711, -0.04782519,  0.01603404, -0.04538268,\n",
+       "         0.00338098, -0.06986066, -0.05395231,  0.12297566,  0.04917476],\n",
+       "       dtype=float32)]"
+      ]
+     },
+     "execution_count": 11,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "model.get_weights()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "1d97bde9-9dcc-46cd-a078-f26f1c276e87",
+   "metadata": {},
+   "source": [
+    "## 8. Modell evaluieren"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "id": "5c224092-5868-4718-8921-bab572acf683",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 2ms/step - accuracy: 0.9749 - loss: 0.0818 \n",
+      "Testgenauigkeit: 0.978600025177002\n",
+      "Test Loss: 0.06802631169557571\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Evaluation\n",
+    "test_loss, test_acc = model.evaluate(x_test, y_test)\n",
+    "print(f\"Testgenauigkeit: {test_acc}\")\n",
+    "print(f\"Test Loss: {test_loss}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "646cbc80-31a8-4e7d-9776-8682ebec0752",
+   "metadata": {},
+   "source": [
+    "## 9. Beispielhafte Vorhersage\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "id": "7e9759ce-76ae-49dc-ab07-e400c18da124",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 2ms/step  \n"
+     ]
+    },
+    {
+     "data": {
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 640x480 with 1 Axes>"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "predictions = model.predict(x_test)\n",
+    "plt.imshow(x_test[0].reshape(28, 28), cmap=\"gray\")\n",
+    "plt.title(f\"Vorhergesagte Klasse: {predictions[0].argmax()}\")\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "010a3735-c35e-4084-9f5e-5c0088e8d7e6",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/docs/2deep_ml_ops/cnn-beispiel.rst b/docs/2deep_ml_ops/cnn-beispiel.rst
deleted file mode 100644
index 4636807..0000000
--- a/docs/2deep_ml_ops/cnn-beispiel.rst
+++ /dev/null
@@ -1,57 +0,0 @@
-Praxisbeispiel – Bildklassifikation mit CNNs
-------------------------------------------------------------
-
-**Ziel:** Einführung in Convolutional Neural Networks mit TensorFlow/Keras anhand eines Bildklassifikationsproblems.
-
-.. note:
-   Code Beispiel auch mit pytorch! als "Gegenbeispiel" zu tensorflows
-
-
-**Code-Beispiel: Klassifikation des MNIST-Datensatzes (Handgeschriebene Ziffern)**
-
-.. code-block:: python
-
-   import tensorflow as tf
-   from tensorflow.keras import layers, models
-   from tensorflow.keras.datasets import mnist
-   import matplotlib.pyplot as plt
-
-   # Daten laden
-   (x_train, y_train), (x_test, y_test) = mnist.load_data()
-   x_train, x_test = x_train / 255.0, x_test / 255.0  # Normalisierung
-   x_train = x_train[..., tf.newaxis]  # Dimension erweitern
-   x_test = x_test[..., tf.newaxis]
-
-   # Modell definieren
-   model = models.Sequential(
-       [
-           layers.Conv2D(32, (3, 3), activation="relu", input_shape=(28, 28, 1)),
-           layers.MaxPooling2D((2, 2)),
-           layers.Conv2D(64, (3, 3), activation="relu"),
-           layers.MaxPooling2D((2, 2)),
-           layers.Conv2D(64, (3, 3), activation="relu"),
-           layers.Flatten(),
-           layers.Dense(64, activation="relu"),
-           layers.Dense(10, activation="softmax"),
-       ]
-   )
-
-   # Modell kompilieren
-   model.compile(
-       optimizer="adam",
-       loss="sparse_categorical_crossentropy",
-       metrics=["accuracy"],
-   )
-
-   # Training
-   model.fit(x_train, y_train, epochs=5, validation_data=(x_test, y_test))
-
-   # Evaluation
-   test_loss, test_acc = model.evaluate(x_test, y_test)
-   print(f"Testgenauigkeit: {test_acc}")
-
-   # Beispielhafte Vorhersage
-   predictions = model.predict(x_test)
-   plt.imshow(x_test[0].reshape(28, 28), cmap="gray")
-   plt.title(f"Vorhergesagte Klasse: {predictions[0].argmax()}")
-   plt.show()
diff --git a/docs/2deep_ml_ops/deduplicate.ipynb b/docs/2deep_ml_ops/deduplicate.ipynb
new file mode 100644
index 0000000..bd21506
--- /dev/null
+++ b/docs/2deep_ml_ops/deduplicate.ipynb
@@ -0,0 +1,1563 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "da945f8f-ac13-4a5d-85e2-76618c5861f3",
+   "metadata": {},
+   "source": [
+    "# Praxisbeispiel - Datenverarbeitung - Deduplizieren\n",
+    "\n",
+    "## 1. Daten \n",
+    "## 1.1 Beispieldaten laden"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "b40d877a-7429-463b-9443-1f423d36121a",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas as pd"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "c5fca3df-d2c0-47a5-81e7-1ffa864bb90a",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "customers = pd.read_csv(\n",
+    "    \"https://raw.githubusercontent.com/kjam/data-cleaning-101/master/data/customer_data_duped.csv\",\n",
+    "    encoding=\"utf-8\",\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "c7bd864d-3e9d-4711-8003-2cd2dd4596b3",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
+  {
+   "cell_type": "markdown",
+   "id": "67b62ec8-4c31-4e77-b46d-32364b01a79b",
+   "metadata": {},
+   "source": [
+    "## 1.2 Daten verstehen \n",
+    "### 1.2.1. Daten anzeigen"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "c7cf08e8-bebd-47f7-824d-60f16c86c85b",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>name</th>\n",
+       "      <th>job</th>\n",
+       "      <th>company</th>\n",
+       "      <th>street_address</th>\n",
+       "      <th>city</th>\n",
+       "      <th>state</th>\n",
+       "      <th>email</th>\n",
+       "      <th>user_name</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>0</th>\n",
+       "      <td>Patricia Schaefer</td>\n",
+       "      <td>Programmer, systems</td>\n",
+       "      <td>Estrada-Best</td>\n",
+       "      <td>398 Paul Drive</td>\n",
+       "      <td>Christianview</td>\n",
+       "      <td>Delaware</td>\n",
+       "      <td>lambdavid@gmail.com</td>\n",
+       "      <td>ndavidson</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1</th>\n",
+       "      <td>Olivie Dubois</td>\n",
+       "      <td>Ingénieur recherche et développement en agroal...</td>\n",
+       "      <td>Moreno</td>\n",
+       "      <td>rue Lucas Benard</td>\n",
+       "      <td>Saint Anastasie-les-Bains</td>\n",
+       "      <td>AR</td>\n",
+       "      <td>berthelotjacqueline@mahe.fr</td>\n",
+       "      <td>manonallain</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2</th>\n",
+       "      <td>Mary Davies-Kirk</td>\n",
+       "      <td>Public affairs consultant</td>\n",
+       "      <td>Baker Ltd</td>\n",
+       "      <td>Flat 3\\nPugh mews</td>\n",
+       "      <td>Stanleyfurt</td>\n",
+       "      <td>ZA</td>\n",
+       "      <td>middletonconor@hotmail.com</td>\n",
+       "      <td>colemanmichael</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>3</th>\n",
+       "      <td>Miroslawa Eckbauer</td>\n",
+       "      <td>Dispensing optician</td>\n",
+       "      <td>Ladeck GmbH</td>\n",
+       "      <td>Mijo-Lübs-Straße 12</td>\n",
+       "      <td>Neubrandenburg</td>\n",
+       "      <td>Berlin</td>\n",
+       "      <td>sophia01@yahoo.de</td>\n",
+       "      <td>romanjunitz</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4</th>\n",
+       "      <td>Richard Bauer</td>\n",
+       "      <td>Accountant, chartered certified</td>\n",
+       "      <td>Hoffman-Rocha</td>\n",
+       "      <td>6541 Rodriguez Wall</td>\n",
+       "      <td>Carlosmouth</td>\n",
+       "      <td>Texas</td>\n",
+       "      <td>tross@jensen-ware.org</td>\n",
+       "      <td>adam78</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>...</th>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2075</th>\n",
+       "      <td>Maurice Stey</td>\n",
+       "      <td>Systems developer</td>\n",
+       "      <td>Linke Margraf GmbH &amp; Co. OHG</td>\n",
+       "      <td>Laila-Scheibe-Allee 2/0</td>\n",
+       "      <td>Luckenwalde</td>\n",
+       "      <td>Hamburg</td>\n",
+       "      <td>gutknechtevelyn@niemeier.com</td>\n",
+       "      <td>dkreusel</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2076</th>\n",
+       "      <td>Linda Alexander</td>\n",
+       "      <td>Commrcil horiculuri</td>\n",
+       "      <td>Webb, Ballald and Vasquel</td>\n",
+       "      <td>5594 Persn Ciff</td>\n",
+       "      <td>Mooneybury</td>\n",
+       "      <td>Maryland</td>\n",
+       "      <td>ahleythoa@ail.co</td>\n",
+       "      <td>kennethrchn</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2077</th>\n",
+       "      <td>Diane Bailly</td>\n",
+       "      <td>Pharmacien</td>\n",
+       "      <td>Voisin</td>\n",
+       "      <td>527, rue Dijoux</td>\n",
+       "      <td>Duval-les-Bains</td>\n",
+       "      <td>CH</td>\n",
+       "      <td>aruiz@reynaud.fr</td>\n",
+       "      <td>dorothee41</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2078</th>\n",
+       "      <td>Jorge Riba Cerdán</td>\n",
+       "      <td>Hotel manager</td>\n",
+       "      <td>Amador-Diego</td>\n",
+       "      <td>Rambla de Adriana Barceló 854 Puerta 3</td>\n",
+       "      <td>Huesca</td>\n",
+       "      <td>Asturias</td>\n",
+       "      <td>manuelamosquera@yahoo.com</td>\n",
+       "      <td>eugenia17</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2079</th>\n",
+       "      <td>Ryan Thompson</td>\n",
+       "      <td>Brewing technologist</td>\n",
+       "      <td>Smith-Sullivan</td>\n",
+       "      <td>136 Rodriguez Point</td>\n",
+       "      <td>Bradfordborough</td>\n",
+       "      <td>North Dakota</td>\n",
+       "      <td>lcruz@gmail.com</td>\n",
+       "      <td>cnewton</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "<p>2080 rows × 8 columns</p>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "                    name                                                job  \\\n",
+       "0      Patricia Schaefer                                Programmer, systems   \n",
+       "1          Olivie Dubois  Ingénieur recherche et développement en agroal...   \n",
+       "2       Mary Davies-Kirk                          Public affairs consultant   \n",
+       "3     Miroslawa Eckbauer                                Dispensing optician   \n",
+       "4          Richard Bauer                    Accountant, chartered certified   \n",
+       "...                  ...                                                ...   \n",
+       "2075        Maurice Stey                                  Systems developer   \n",
+       "2076     Linda Alexander                                Commrcil horiculuri   \n",
+       "2077        Diane Bailly                                         Pharmacien   \n",
+       "2078   Jorge Riba Cerdán                                      Hotel manager   \n",
+       "2079       Ryan Thompson                               Brewing technologist   \n",
+       "\n",
+       "                           company                           street_address  \\\n",
+       "0                     Estrada-Best                           398 Paul Drive   \n",
+       "1                           Moreno                         rue Lucas Benard   \n",
+       "2                        Baker Ltd                        Flat 3\\nPugh mews   \n",
+       "3                      Ladeck GmbH                      Mijo-Lübs-Straße 12   \n",
+       "4                    Hoffman-Rocha                      6541 Rodriguez Wall   \n",
+       "...                            ...                                      ...   \n",
+       "2075  Linke Margraf GmbH & Co. OHG                  Laila-Scheibe-Allee 2/0   \n",
+       "2076     Webb, Ballald and Vasquel                          5594 Persn Ciff   \n",
+       "2077                        Voisin                          527, rue Dijoux   \n",
+       "2078                  Amador-Diego  Rambla de Adriana Barceló 854 Puerta 3    \n",
+       "2079                Smith-Sullivan                      136 Rodriguez Point   \n",
+       "\n",
+       "                           city         state                         email  \\\n",
+       "0                 Christianview      Delaware           lambdavid@gmail.com   \n",
+       "1     Saint Anastasie-les-Bains            AR   berthelotjacqueline@mahe.fr   \n",
+       "2                   Stanleyfurt            ZA    middletonconor@hotmail.com   \n",
+       "3                Neubrandenburg        Berlin             sophia01@yahoo.de   \n",
+       "4                   Carlosmouth         Texas         tross@jensen-ware.org   \n",
+       "...                         ...           ...                           ...   \n",
+       "2075                Luckenwalde       Hamburg  gutknechtevelyn@niemeier.com   \n",
+       "2076                 Mooneybury      Maryland              ahleythoa@ail.co   \n",
+       "2077            Duval-les-Bains            CH              aruiz@reynaud.fr   \n",
+       "2078                     Huesca      Asturias     manuelamosquera@yahoo.com   \n",
+       "2079            Bradfordborough  North Dakota               lcruz@gmail.com   \n",
+       "\n",
+       "           user_name  \n",
+       "0          ndavidson  \n",
+       "1        manonallain  \n",
+       "2     colemanmichael  \n",
+       "3        romanjunitz  \n",
+       "4             adam78  \n",
+       "...              ...  \n",
+       "2075        dkreusel  \n",
+       "2076     kennethrchn  \n",
+       "2077      dorothee41  \n",
+       "2078       eugenia17  \n",
+       "2079         cnewton  \n",
+       "\n",
+       "[2080 rows x 8 columns]"
+      ]
+     },
+     "execution_count": 3,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "customers"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "c4c0955c-a8cc-4fbe-b304-cecc97f41bea",
+   "metadata": {},
+   "source": [
+    "### 1.2.2. Datentypen anzeigen \n",
+    "Hierfür verwenden wir pandas.DataFrame.dtypes:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "578db81c-f9ad-47c1-bf1f-c4183e6498f4",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "name              object\n",
+       "job               object\n",
+       "company           object\n",
+       "street_address    object\n",
+       "city              object\n",
+       "state             object\n",
+       "email             object\n",
+       "user_name         object\n",
+       "dtype: object"
+      ]
+     },
+     "execution_count": 5,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "customers.dtypes\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "c58db531-b981-43a2-ac1b-319eb321c767",
+   "metadata": {},
+   "source": [
+    "### 1.2.3. Fehlende Werte ermitteln\n",
+    "pandas.isnull zeigt für ein array-ähnliches Objekt an, ob Werte fehlen:\n",
+    "\n",
+    "- NaN in numerischen Arrays\n",
+    "\n",
+    "- None oder NaN in Objekt-Arrays\n",
+    "\n",
+    "- NaT in datetimelike\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "0540e3f9-4db0-4b91-9f7f-878f53dc1a44",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "name 0\n",
+      "job 0\n",
+      "company 0\n",
+      "street_address 0\n",
+      "city 0\n",
+      "state 0\n",
+      "email 0\n",
+      "user_name 0\n"
+     ]
+    }
+   ],
+   "source": [
+    "for col in customers.columns:\n",
+    "    print(col, customers[col].isnull().sum())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "799d56c7-a0e3-46a3-943c-befaacc5a74d",
+   "metadata": {},
+   "source": [
+    "## 2. Duplizierte Datensätze ermitteln und löschen\n",
+    "### 2.1. mit Pandas "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "a53120ae-2f8a-4bb5-b7a3-4cb3e96f55c0",
+   "metadata": {},
+   "source": [
+    "#### 2.1.1 Duplizierte Datenpunkte ermitteln"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "1ae2365b-d980-409f-9f36-e2cee7525a35",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "0       False\n",
+       "1       False\n",
+       "2       False\n",
+       "3       False\n",
+       "4       False\n",
+       "        ...  \n",
+       "2075    False\n",
+       "2076    False\n",
+       "2077    False\n",
+       "2078    False\n",
+       "2079    False\n",
+       "Length: 2080, dtype: bool"
+      ]
+     },
+     "execution_count": 8,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "customers.duplicated()\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "30a1a1c0-839d-458c-8cce-853895cff9a8",
+   "metadata": {},
+   "source": [
+    "customers.duplicated() gibt uns noch nicht den gewünschten Hinweis, ob es doppelte Datensätze gibt. Im Folgenden lassen wir uns alle Datensätze ausgeben, für die True zurückgegeben wird:\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "8857b4f6-ec55-4c24-8adc-1645eca3c1c1",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>name</th>\n",
+       "      <th>job</th>\n",
+       "      <th>company</th>\n",
+       "      <th>street_address</th>\n",
+       "      <th>city</th>\n",
+       "      <th>state</th>\n",
+       "      <th>email</th>\n",
+       "      <th>user_name</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "Empty DataFrame\n",
+       "Columns: [name, job, company, street_address, city, state, email, user_name]\n",
+       "Index: []"
+      ]
+     },
+     "execution_count": 9,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "customers[customers.duplicated()]\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "5b07a4fd-a06c-4d6e-8b31-9600ff5b5726",
+   "metadata": {},
+   "source": [
+    "Offenbar gibt es keine identischen Datensätze.\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "bd1c85e5-526b-4831-abba-77cab5d0d095",
+   "metadata": {},
+   "source": [
+    "### 2.1.2. Deduplizieren \n",
+    "\n",
+    "Da pandas keine Duplikate erkennen konnte, wird das Löschen doppelter Datensätzemit drop_duplicates demnach nichts ändern und die Anzahl des Datensatzes bei 2080 belassen.\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "757fa2fb-381c-4a92-b001-6ddf0ea51d7e",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "(2080, 8)\n",
+      "(2080, 8)\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(customers.shape)\n",
+    "customers.drop_duplicates(inplace=True)\n",
+    "print(customers.shape)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "349b31db-dbe8-4fd1-b7a1-a6dea4192c04",
+   "metadata": {},
+   "source": [
+    "### 2.1.3. user_name-spezifische Ermittlung und Deduplizierung"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "a219298d-9998-4540-9b9e-a7af85b4d47f",
+   "metadata": {},
+   "source": [
+    "Wir gehen einen Detailschritt weiter und lassen uns diejenigen Datensätze anzeigen, bei denen nur der user_name identisch ist:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "id": "950221bf-7983-4bc7-bb80-8f250b6903fe",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>name</th>\n",
+       "      <th>job</th>\n",
+       "      <th>company</th>\n",
+       "      <th>street_address</th>\n",
+       "      <th>city</th>\n",
+       "      <th>state</th>\n",
+       "      <th>email</th>\n",
+       "      <th>user_name</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>337</th>\n",
+       "      <td>Aysel Binner</td>\n",
+       "      <td>Reccig officer</td>\n",
+       "      <td>Kuhl Kalleww Swifwunw &amp; Co. KGaA</td>\n",
+       "      <td>Batix-Kanz-Staß 5/4</td>\n",
+       "      <td>Fulda</td>\n",
+       "      <td>Berli</td>\n",
+       "      <td>frncoise@wgnerco</td>\n",
+       "      <td>christinefinke</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>377</th>\n",
+       "      <td>Jolanta Rogge</td>\n",
+       "      <td>Accommodation managr</td>\n",
+       "      <td>Scholl e.V.</td>\n",
+       "      <td>Lrchplz 4/6</td>\n",
+       "      <td>Mettmnn</td>\n",
+       "      <td>Thüringen</td>\n",
+       "      <td>inrharff@yah.d</td>\n",
+       "      <td>walentinabeier</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>506</th>\n",
+       "      <td>Mrs. Frances Peters</td>\n",
+       "      <td>Fuiue desie</td>\n",
+       "      <td>Rsgers, Lawrence and Richards</td>\n",
+       "      <td>Studio \\nCarpntr kys</td>\n",
+       "      <td>Wes Simn</td>\n",
+       "      <td>BO</td>\n",
+       "      <td>halenewilliams@wilson-sandes.og</td>\n",
+       "      <td>amy17</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>545</th>\n",
+       "      <td>Gerhart Krebs MBA.</td>\n",
+       "      <td>Surgeon</td>\n",
+       "      <td>Roskoth</td>\n",
+       "      <td>Kühnertweg 863</td>\n",
+       "      <td>Stade</td>\n",
+       "      <td>Bayern</td>\n",
+       "      <td>olav44@bolander.de</td>\n",
+       "      <td>bettyhahn</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>592</th>\n",
+       "      <td>Folkert Gnatz</td>\n",
+       "      <td>Meteorologist</td>\n",
+       "      <td>Bolnbach</td>\n",
+       "      <td>Heinfried-Austermühle-Ring 05</td>\n",
+       "      <td>Eilenburg</td>\n",
+       "      <td>Thüringen</td>\n",
+       "      <td>jaentschbirgitt@boerner.org</td>\n",
+       "      <td>francesco44</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>633</th>\n",
+       "      <td>Manon Jacquot</td>\n",
+       "      <td>Ingénieur en aéronautique</td>\n",
+       "      <td>Jacob</td>\n",
+       "      <td>8, chemin Éléonore Evrard</td>\n",
+       "      <td>Marechal-les-Bains</td>\n",
+       "      <td>AR</td>\n",
+       "      <td>ilemaitre@voila.fr</td>\n",
+       "      <td>astrid58</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>658</th>\n",
+       "      <td>Austin Waller</td>\n",
+       "      <td>Insurance risk surveyor</td>\n",
+       "      <td>Sexton Group</td>\n",
+       "      <td>11097 Hansen Field</td>\n",
+       "      <td>Davidmouth</td>\n",
+       "      <td>Texas</td>\n",
+       "      <td>christina74@doyle-baker.biz</td>\n",
+       "      <td>olynn</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>723</th>\n",
+       "      <td>Wanda Moran</td>\n",
+       "      <td>Solicitor, Scotland</td>\n",
+       "      <td>Estes PLC</td>\n",
+       "      <td>08011 Hernandez Streets Apt. 149</td>\n",
+       "      <td>Natalieshire</td>\n",
+       "      <td>Oregon</td>\n",
+       "      <td>howardreginald@gmail.com</td>\n",
+       "      <td>dana91</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>762</th>\n",
+       "      <td>Charles Russell</td>\n",
+       "      <td>Scientist, research (physical sciences)</td>\n",
+       "      <td>Preston-Wilson</td>\n",
+       "      <td>6709 Ashley Circle Apt. 309</td>\n",
+       "      <td>Danielberg</td>\n",
+       "      <td>South Dakota</td>\n",
+       "      <td>nancyescobar@brown.net</td>\n",
+       "      <td>ruben71</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>772</th>\n",
+       "      <td>Waltrud Wohlgemut</td>\n",
+       "      <td>Designer, fashion/clothing</td>\n",
+       "      <td>Nerger AG</td>\n",
+       "      <td>Elmar-Ullmann-Allee 6</td>\n",
+       "      <td>Schlüchtern</td>\n",
+       "      <td>Rheinland-Pfalz</td>\n",
+       "      <td>auch-schlauchindietlind@gmx.de</td>\n",
+       "      <td>zitakuhl</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>783</th>\n",
+       "      <td>Caroline Mata</td>\n",
+       "      <td>Engineer, elecrical</td>\n",
+       "      <td>Grimes Grrur</td>\n",
+       "      <td>80157 Whte Alley Sute 79</td>\n",
+       "      <td>Soh Mark</td>\n",
+       "      <td>Iw</td>\n",
+       "      <td>jared52@aoo.com</td>\n",
+       "      <td>thomasthompson</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>889</th>\n",
+       "      <td>Ricardo Ripoll Lucena</td>\n",
+       "      <td>Teevisi camera peratr</td>\n",
+       "      <td>Luzq Estraqa anq Galinqq</td>\n",
+       "      <td>Caejón Rosario Viapana 16</td>\n",
+       "      <td>Palencia</td>\n",
+       "      <td>Lgo</td>\n",
+       "      <td>ev0@oo.com</td>\n",
+       "      <td>colomerenrique</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>928</th>\n",
+       "      <td>Sophie Letellier du Carpentier</td>\n",
+       "      <td>Cnucteu e ét</td>\n",
+       "      <td>Valle7 SARL</td>\n",
+       "      <td>3, boulvard Jan Augr</td>\n",
+       "      <td>Saint Daviddan</td>\n",
+       "      <td>BS</td>\n",
+       "      <td>rdorm@dbmi.com</td>\n",
+       "      <td>anne28</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>979</th>\n",
+       "      <td>Irene Roda Dávila</td>\n",
+       "      <td>Eitor, maazine featres</td>\n",
+       "      <td>Daza Inc</td>\n",
+       "      <td>Roda Carla Miró 5</td>\n",
+       "      <td>Viy</td>\n",
+       "      <td>La Rioa</td>\n",
+       "      <td>sldrpére@ps.cm</td>\n",
+       "      <td>ipeñalver</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>995</th>\n",
+       "      <td>Abigail Hernandez</td>\n",
+       "      <td>Mechanical engineer</td>\n",
+       "      <td>Smith Ltd</td>\n",
+       "      <td>766 Adrian Ranch</td>\n",
+       "      <td>Ellismouth</td>\n",
+       "      <td>Colorado</td>\n",
+       "      <td>jordan60@gmail.com</td>\n",
+       "      <td>mendozajody</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1015</th>\n",
+       "      <td>Mr. Paul Newton</td>\n",
+       "      <td>Government soa researh offer</td>\n",
+       "      <td>LemnardmWatsmn</td>\n",
+       "      <td>Studi 86\\nKaty ill</td>\n",
+       "      <td>West Jue</td>\n",
+       "      <td>VE</td>\n",
+       "      <td>em@mil.cm</td>\n",
+       "      <td>bbennett</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1043</th>\n",
+       "      <td>Anna Adams</td>\n",
+       "      <td>Programmer alcatons</td>\n",
+       "      <td>Jones Gjoup</td>\n",
+       "      <td>22 Kateen ova</td>\n",
+       "      <td>Noth Joa</td>\n",
+       "      <td>KZ</td>\n",
+       "      <td>asleig65@aisay.co</td>\n",
+       "      <td>lloydann</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1052</th>\n",
+       "      <td>Aurélie Vidal</td>\n",
+       "      <td>Magistrat</td>\n",
+       "      <td>Martins</td>\n",
+       "      <td>88, rue Stéphanie Letellier</td>\n",
+       "      <td>Rouxnec</td>\n",
+       "      <td>SE</td>\n",
+       "      <td>boutineric@blin.fr</td>\n",
+       "      <td>iwagner</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1062</th>\n",
+       "      <td>Regina Schacht-Kusch</td>\n",
+       "      <td>Herbalist</td>\n",
+       "      <td>Hartung GmbH &amp; Co. KGaA</td>\n",
+       "      <td>Wenke-Hörle-Ring 36</td>\n",
+       "      <td>Eggenfelden</td>\n",
+       "      <td>Sachsen-Anhalt</td>\n",
+       "      <td>oluebs@troest.de</td>\n",
+       "      <td>xklotz</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1120</th>\n",
+       "      <td>Jeffrey Benjamin</td>\n",
+       "      <td>Publ house manager</td>\n",
+       "      <td>Chcn Inc</td>\n",
+       "      <td>27 Rodgrs Rdgs Apt. 269</td>\n",
+       "      <td>Suth Jeffererg</td>\n",
+       "      <td>Iinois</td>\n",
+       "      <td>stepanie90@rogers.co</td>\n",
+       "      <td>lori67</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1170</th>\n",
+       "      <td>Julio Agustín Amaya</td>\n",
+       "      <td>Tax aviser</td>\n",
+       "      <td>Piñolk Belmonke and Codina</td>\n",
+       "      <td>Calleón de Gregorio Bustamante 28 Piso 7</td>\n",
+       "      <td>La Pala</td>\n",
+       "      <td>Salamanca</td>\n",
+       "      <td>usolana@jáuregui-pedraza.om</td>\n",
+       "      <td>gloriaolmo</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1339</th>\n",
+       "      <td>Ing. Andrew Schleich B.A.</td>\n",
+       "      <td>Ln</td>\n",
+       "      <td>Holt Putz GnR</td>\n",
+       "      <td>Hugasse 8/8</td>\n",
+       "      <td>Hainichn</td>\n",
+       "      <td>Neersachsen</td>\n",
+       "      <td>jun@putz.com</td>\n",
+       "      <td>jesselmaja</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1360</th>\n",
+       "      <td>Frédérique Lejeune-Daniel</td>\n",
+       "      <td>Tecce cse</td>\n",
+       "      <td>Sctmitt</td>\n",
+       "      <td>chemin Denise Ferrand</td>\n",
+       "      <td>Saint ChalotteVille</td>\n",
+       "      <td>IE</td>\n",
+       "      <td>jchretien@costacom</td>\n",
+       "      <td>joseph60</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1384</th>\n",
+       "      <td>Kenneth Moore</td>\n",
+       "      <td>Magazine journalist</td>\n",
+       "      <td>Cross, Bfll anf Diaz</td>\n",
+       "      <td>753 Lindsey Pine</td>\n",
+       "      <td>Thompsonshe</td>\n",
+       "      <td>Colorao</td>\n",
+       "      <td>ashey28@rice.co</td>\n",
+       "      <td>todd72</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1423</th>\n",
+       "      <td>Thomas Coulon</td>\n",
+       "      <td>Collecteur de fonds</td>\n",
+       "      <td>Levy</td>\n",
+       "      <td>91, rue Laetitia Collet</td>\n",
+       "      <td>Dias-sur-Normand</td>\n",
+       "      <td>SC</td>\n",
+       "      <td>deschampsgabriel@guyot.fr</td>\n",
+       "      <td>michelepetit</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1433</th>\n",
+       "      <td>Jerry Barnes</td>\n",
+       "      <td>Tour mner</td>\n",
+       "      <td>Col-Wllllams</td>\n",
+       "      <td>30 Mpy Ovepass</td>\n",
+       "      <td>Jeiferview</td>\n",
+       "      <td>Utah</td>\n",
+       "      <td>insnashl@gas-hais.cm</td>\n",
+       "      <td>christopher62</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1452</th>\n",
+       "      <td>Karen Weeks</td>\n",
+       "      <td>Psychotherapist, child</td>\n",
+       "      <td>Rodriguez, Brady and Jackson</td>\n",
+       "      <td>233 Kevin Street</td>\n",
+       "      <td>Larryside</td>\n",
+       "      <td>Indiana</td>\n",
+       "      <td>gregg39@hernandez-gomez.com</td>\n",
+       "      <td>knapprobert</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1489</th>\n",
+       "      <td>Herr Johann Eigenwillig</td>\n",
+       "      <td>Immigration officer</td>\n",
+       "      <td>Süßebier Hänel GmbH</td>\n",
+       "      <td>Langernplatz 0</td>\n",
+       "      <td>Stadtsteinach</td>\n",
+       "      <td>Thüringen</td>\n",
+       "      <td>haasemarieluise@noack.com</td>\n",
+       "      <td>istoll</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1544</th>\n",
+       "      <td>Pasquale Schwital</td>\n",
+       "      <td>Trade mark attorney</td>\n",
+       "      <td>Finke</td>\n",
+       "      <td>Detlef-Binner-Platz 0/1</td>\n",
+       "      <td>Burg</td>\n",
+       "      <td>Niedersachsen</td>\n",
+       "      <td>hanne-lore98@gmx.de</td>\n",
+       "      <td>thomas14</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1557</th>\n",
+       "      <td>Stephanie Young</td>\n",
+       "      <td>Herpetologist</td>\n",
+       "      <td>Bryant and Sons</td>\n",
+       "      <td>5163 Rebecca Creek Suite 421</td>\n",
+       "      <td>North Theresaberg</td>\n",
+       "      <td>Alaska</td>\n",
+       "      <td>stephenwilliams@summers.com</td>\n",
+       "      <td>ahawkins</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1567</th>\n",
+       "      <td>Carolina Reguera Sanz</td>\n",
+       "      <td>Fam manae</td>\n",
+       "      <td>Cami77, C7aparr7 a7d N7gu7ra</td>\n",
+       "      <td>Vil e Imel Oorio 25</td>\n",
+       "      <td>Madd</td>\n",
+       "      <td>Vicaya</td>\n",
+       "      <td>mordóñ@cámara.info</td>\n",
+       "      <td>eva16</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1616</th>\n",
+       "      <td>Sonia Amores</td>\n",
+       "      <td>Senir tax prfessina/tax inspectr</td>\n",
+       "      <td>J5an-Núñez</td>\n",
+       "      <td>Avnida d Grgorio Manón 344 Prta 8</td>\n",
+       "      <td>Ponevedr</td>\n",
+       "      <td>Lugo</td>\n",
+       "      <td>icent4@montenero-brroso.info</td>\n",
+       "      <td>sanmartínguillermo</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1647</th>\n",
+       "      <td>Juan Carlos Iker Boix Ros</td>\n",
+       "      <td>Pre phtgrapher</td>\n",
+       "      <td>Pont, P44om4r4s 4nd Arjon4</td>\n",
+       "      <td>Pasadzo de Josep Bentez  Pso</td>\n",
+       "      <td>Las Palmas</td>\n",
+       "      <td>Mia</td>\n",
+       "      <td>srgio24@gail.co</td>\n",
+       "      <td>luis-miguel23</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1652</th>\n",
+       "      <td>Jörg Henschel</td>\n",
+       "      <td>Chaity office</td>\n",
+       "      <td>Schicke AG</td>\n",
+       "      <td>HennyLorchRng 484</td>\n",
+       "      <td>Hohensein-Ensh</td>\n",
+       "      <td>BadenWürtteberg</td>\n",
+       "      <td>huerhes@hmal.de</td>\n",
+       "      <td>anne-katrin51</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1703</th>\n",
+       "      <td>Marc Tate</td>\n",
+       "      <td>Ship broker</td>\n",
+       "      <td>Wagner, Mitchell and Grimes</td>\n",
+       "      <td>721 Christopher View Suite 840</td>\n",
+       "      <td>Watsonmouth</td>\n",
+       "      <td>Connecticut</td>\n",
+       "      <td>chenjessica@hotmail.com</td>\n",
+       "      <td>patricia34</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1707</th>\n",
+       "      <td>Joseph Hines</td>\n",
+       "      <td>Pyhiatri nre</td>\n",
+       "      <td>Cr4ig, G4rci4 4nd Rich4rds</td>\n",
+       "      <td>85663 Savage Gles</td>\n",
+       "      <td>Mcgeeon</td>\n",
+       "      <td>Als</td>\n",
+       "      <td>bcaldern@htmail.cm</td>\n",
+       "      <td>emilytorres</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1722</th>\n",
+       "      <td>Julie Baldwin</td>\n",
+       "      <td>Set deigner</td>\n",
+       "      <td>W5ll55mson-G5rz5</td>\n",
+       "      <td>58513 Paricia Res Suie 45</td>\n",
+       "      <td>So Me</td>\n",
+       "      <td>Alaska</td>\n",
+       "      <td>diuez@uess.</td>\n",
+       "      <td>cmoss</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1759</th>\n",
+       "      <td>Sarah Hoffman</td>\n",
+       "      <td>Exhibitin designe</td>\n",
+       "      <td>Hensont Wiley and Ryan</td>\n",
+       "      <td>9490 Curts Spur Sute 82</td>\n",
+       "      <td>Jseptwn</td>\n",
+       "      <td>Arizona</td>\n",
+       "      <td>ncole@yahoo.com</td>\n",
+       "      <td>csmith</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1796</th>\n",
+       "      <td>Valentine Devaux-Roger</td>\n",
+       "      <td>Direceur d'ôial</td>\n",
+       "      <td>Leiris</td>\n",
+       "      <td>57, enue de Gros</td>\n",
+       "      <td>BenadBou</td>\n",
+       "      <td>AL</td>\n",
+       "      <td>rogrlro@munoz.om</td>\n",
+       "      <td>xherve</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1809</th>\n",
+       "      <td>Slavica Seidel</td>\n",
+       "      <td>Psychotherapist, child</td>\n",
+       "      <td>Wulff Hande KG</td>\n",
+       "      <td>Preißgasse 0/4</td>\n",
+       "      <td>Soest</td>\n",
+       "      <td>Rheinland-Pfalz</td>\n",
+       "      <td>tloos@krause.net</td>\n",
+       "      <td>abien</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1820</th>\n",
+       "      <td>Wenke Schweitzer</td>\n",
+       "      <td>Enginr, automoti</td>\n",
+       "      <td>Wesa4k KG</td>\n",
+       "      <td>Eies. 7</td>\n",
+       "      <td>Ba Lnwra</td>\n",
+       "      <td>Thürige</td>\n",
+       "      <td>rsthveriue@mies.rg</td>\n",
+       "      <td>kwernecke</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1829</th>\n",
+       "      <td>Dr. Thomas Hein</td>\n",
+       "      <td>Copy</td>\n",
+       "      <td>Geisel</td>\n",
+       "      <td>Ladeckgasse 11</td>\n",
+       "      <td>Rockenhausen</td>\n",
+       "      <td>Nordrhein-Westfalen</td>\n",
+       "      <td>grein-grotharnim@kallert.de</td>\n",
+       "      <td>siegmar08</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1837</th>\n",
+       "      <td>Andrew Hart</td>\n",
+       "      <td>Engineer, civil (contracting)</td>\n",
+       "      <td>Barnett LLC</td>\n",
+       "      <td>258 Day Hollow Suite 410</td>\n",
+       "      <td>Kimberlyhaven</td>\n",
+       "      <td>Colorado</td>\n",
+       "      <td>brandy00@yahoo.com</td>\n",
+       "      <td>amy30</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1914</th>\n",
+       "      <td>Shelby Fowler</td>\n",
+       "      <td>Air traffic controller</td>\n",
+       "      <td>Fields-Sanchez</td>\n",
+       "      <td>533 Fitzpatrick Bypass</td>\n",
+       "      <td>Francesberg</td>\n",
+       "      <td>Michigan</td>\n",
+       "      <td>terrystephen@anderson.org</td>\n",
+       "      <td>gcain</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1938</th>\n",
+       "      <td>Susan Aubry</td>\n",
+       "      <td>Directeur d'agence bancaire</td>\n",
+       "      <td>Payet Georges S.A.S.</td>\n",
+       "      <td>67, rue Inès Valentin</td>\n",
+       "      <td>Nicolas</td>\n",
+       "      <td>FI</td>\n",
+       "      <td>milletedith@sfr.fr</td>\n",
+       "      <td>tthierry</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1948</th>\n",
+       "      <td>Richard Karge-Kobelt</td>\n",
+       "      <td>Junalist maaine</td>\n",
+       "      <td>Abberb Keubeb AG</td>\n",
+       "      <td>Mitschkeee 8</td>\n",
+       "      <td>Mß</td>\n",
+       "      <td>SachsnAnhalt</td>\n",
+       "      <td>nrejwgner@gmx.e</td>\n",
+       "      <td>muehlehenni</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1960</th>\n",
+       "      <td>Anna de Lobato</td>\n",
+       "      <td>Medcl techcl ocer</td>\n",
+       "      <td>Maciag PLC</td>\n",
+       "      <td>Calleón de Dolore Parea 21 At 7</td>\n",
+       "      <td>Palncia</td>\n",
+       "      <td>Cantaria</td>\n",
+       "      <td>vázqzlornzo@al.om</td>\n",
+       "      <td>daniel70</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1968</th>\n",
+       "      <td>Zoltan Wähner B.A.</td>\n",
+       "      <td>Professor Emerits</td>\n",
+       "      <td>Th8e8</td>\n",
+       "      <td>Stotr. 1</td>\n",
+       "      <td>Saulgau</td>\n",
+       "      <td>Shlsg-Holst</td>\n",
+       "      <td>arlenpruschke@salz.or</td>\n",
+       "      <td>kklemm</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1995</th>\n",
+       "      <td>Kenneth Dunn</td>\n",
+       "      <td>Programmer, systems</td>\n",
+       "      <td>Leonard Inc</td>\n",
+       "      <td>5361 Patterson Mission Suite 504</td>\n",
+       "      <td>Villaburgh</td>\n",
+       "      <td>Rhode Island</td>\n",
+       "      <td>kristen54@gmail.com</td>\n",
+       "      <td>jkent</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2010</th>\n",
+       "      <td>Gertraude Schomber</td>\n",
+       "      <td>Insurance risk surveyor</td>\n",
+       "      <td>Bruder</td>\n",
+       "      <td>Christa-Ullrich-Allee 0/1</td>\n",
+       "      <td>Schwäbisch Hall</td>\n",
+       "      <td>Hessen</td>\n",
+       "      <td>gumprichalice@schmidt.de</td>\n",
+       "      <td>fruppert</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2075</th>\n",
+       "      <td>Maurice Stey</td>\n",
+       "      <td>Systems developer</td>\n",
+       "      <td>Linke Margraf GmbH &amp; Co. OHG</td>\n",
+       "      <td>Laila-Scheibe-Allee 2/0</td>\n",
+       "      <td>Luckenwalde</td>\n",
+       "      <td>Hamburg</td>\n",
+       "      <td>gutknechtevelyn@niemeier.com</td>\n",
+       "      <td>dkreusel</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "                                name                                      job  \\\n",
+       "337                     Aysel Binner                           Reccig officer   \n",
+       "377                    Jolanta Rogge                     Accommodation managr   \n",
+       "506              Mrs. Frances Peters                              Fuiue desie   \n",
+       "545               Gerhart Krebs MBA.                                  Surgeon   \n",
+       "592                    Folkert Gnatz                            Meteorologist   \n",
+       "633                    Manon Jacquot                Ingénieur en aéronautique   \n",
+       "658                    Austin Waller                  Insurance risk surveyor   \n",
+       "723                      Wanda Moran                      Solicitor, Scotland   \n",
+       "762                  Charles Russell  Scientist, research (physical sciences)   \n",
+       "772                Waltrud Wohlgemut               Designer, fashion/clothing   \n",
+       "783                    Caroline Mata                      Engineer, elecrical   \n",
+       "889            Ricardo Ripoll Lucena                    Teevisi camera peratr   \n",
+       "928   Sophie Letellier du Carpentier                             Cnucteu e ét   \n",
+       "979                Irene Roda Dávila                   Eitor, maazine featres   \n",
+       "995                Abigail Hernandez                      Mechanical engineer   \n",
+       "1015                 Mr. Paul Newton             Government soa researh offer   \n",
+       "1043                      Anna Adams                      Programmer alcatons   \n",
+       "1052                   Aurélie Vidal                                Magistrat   \n",
+       "1062            Regina Schacht-Kusch                                Herbalist   \n",
+       "1120                Jeffrey Benjamin                       Publ house manager   \n",
+       "1170             Julio Agustín Amaya                               Tax aviser   \n",
+       "1339       Ing. Andrew Schleich B.A.                                       Ln   \n",
+       "1360       Frédérique Lejeune-Daniel                                Tecce cse   \n",
+       "1384                   Kenneth Moore                      Magazine journalist   \n",
+       "1423                   Thomas Coulon                      Collecteur de fonds   \n",
+       "1433                    Jerry Barnes                                Tour mner   \n",
+       "1452                     Karen Weeks                   Psychotherapist, child   \n",
+       "1489         Herr Johann Eigenwillig                      Immigration officer   \n",
+       "1544               Pasquale Schwital                      Trade mark attorney   \n",
+       "1557                 Stephanie Young                            Herpetologist   \n",
+       "1567           Carolina Reguera Sanz                                Fam manae   \n",
+       "1616                    Sonia Amores         Senir tax prfessina/tax inspectr   \n",
+       "1647       Juan Carlos Iker Boix Ros                           Pre phtgrapher   \n",
+       "1652                   Jörg Henschel                            Chaity office   \n",
+       "1703                       Marc Tate                              Ship broker   \n",
+       "1707                    Joseph Hines                             Pyhiatri nre   \n",
+       "1722                   Julie Baldwin                              Set deigner   \n",
+       "1759                   Sarah Hoffman                        Exhibitin designe   \n",
+       "1796          Valentine Devaux-Roger                          Direceur d'ôial   \n",
+       "1809                  Slavica Seidel                   Psychotherapist, child   \n",
+       "1820                Wenke Schweitzer                         Enginr, automoti   \n",
+       "1829                 Dr. Thomas Hein                                     Copy   \n",
+       "1837                     Andrew Hart            Engineer, civil (contracting)   \n",
+       "1914                   Shelby Fowler                   Air traffic controller   \n",
+       "1938                     Susan Aubry              Directeur d'agence bancaire   \n",
+       "1948            Richard Karge-Kobelt                          Junalist maaine   \n",
+       "1960                  Anna de Lobato                        Medcl techcl ocer   \n",
+       "1968              Zoltan Wähner B.A.                        Professor Emerits   \n",
+       "1995                    Kenneth Dunn                      Programmer, systems   \n",
+       "2010              Gertraude Schomber                  Insurance risk surveyor   \n",
+       "2075                    Maurice Stey                        Systems developer   \n",
+       "\n",
+       "                               company  \\\n",
+       "337   Kuhl Kalleww Swifwunw & Co. KGaA   \n",
+       "377                        Scholl e.V.   \n",
+       "506      Rsgers, Lawrence and Richards   \n",
+       "545                            Roskoth   \n",
+       "592                           Bolnbach   \n",
+       "633                              Jacob   \n",
+       "658                       Sexton Group   \n",
+       "723                          Estes PLC   \n",
+       "762                     Preston-Wilson   \n",
+       "772                          Nerger AG   \n",
+       "783                       Grimes Grrur   \n",
+       "889           Luzq Estraqa anq Galinqq   \n",
+       "928                        Valle7 SARL   \n",
+       "979                           Daza Inc   \n",
+       "995                          Smith Ltd   \n",
+       "1015                    LemnardmWatsmn   \n",
+       "1043                       Jones Gjoup   \n",
+       "1052                           Martins   \n",
+       "1062           Hartung GmbH & Co. KGaA   \n",
+       "1120                          Chcn Inc   \n",
+       "1170        Piñolk Belmonke and Codina   \n",
+       "1339                     Holt Putz GnR   \n",
+       "1360                           Sctmitt   \n",
+       "1384              Cross, Bfll anf Diaz   \n",
+       "1423                              Levy   \n",
+       "1433                      Col-Wllllams   \n",
+       "1452      Rodriguez, Brady and Jackson   \n",
+       "1489               Süßebier Hänel GmbH   \n",
+       "1544                             Finke   \n",
+       "1557                   Bryant and Sons   \n",
+       "1567      Cami77, C7aparr7 a7d N7gu7ra   \n",
+       "1616                        J5an-Núñez   \n",
+       "1647        Pont, P44om4r4s 4nd Arjon4   \n",
+       "1652                        Schicke AG   \n",
+       "1703       Wagner, Mitchell and Grimes   \n",
+       "1707        Cr4ig, G4rci4 4nd Rich4rds   \n",
+       "1722                  W5ll55mson-G5rz5   \n",
+       "1759            Hensont Wiley and Ryan   \n",
+       "1796                            Leiris   \n",
+       "1809                    Wulff Hande KG   \n",
+       "1820                         Wesa4k KG   \n",
+       "1829                            Geisel   \n",
+       "1837                       Barnett LLC   \n",
+       "1914                    Fields-Sanchez   \n",
+       "1938              Payet Georges S.A.S.   \n",
+       "1948                  Abberb Keubeb AG   \n",
+       "1960                        Maciag PLC   \n",
+       "1968                             Th8e8   \n",
+       "1995                       Leonard Inc   \n",
+       "2010                            Bruder   \n",
+       "2075      Linke Margraf GmbH & Co. OHG   \n",
+       "\n",
+       "                                 street_address                 city  \\\n",
+       "337                         Batix-Kanz-Staß 5/4                Fulda   \n",
+       "377                                 Lrchplz 4/6              Mettmnn   \n",
+       "506                        Studio \\nCarpntr kys             Wes Simn   \n",
+       "545                              Kühnertweg 863                Stade   \n",
+       "592               Heinfried-Austermühle-Ring 05            Eilenburg   \n",
+       "633                   8, chemin Éléonore Evrard   Marechal-les-Bains   \n",
+       "658                          11097 Hansen Field           Davidmouth   \n",
+       "723            08011 Hernandez Streets Apt. 149         Natalieshire   \n",
+       "762                 6709 Ashley Circle Apt. 309           Danielberg   \n",
+       "772                       Elmar-Ullmann-Allee 6          Schlüchtern   \n",
+       "783                    80157 Whte Alley Sute 79             Soh Mark   \n",
+       "889                   Caejón Rosario Viapana 16             Palencia   \n",
+       "928                        3, boulvard Jan Augr       Saint Daviddan   \n",
+       "979                           Roda Carla Miró 5                  Viy   \n",
+       "995                            766 Adrian Ranch           Ellismouth   \n",
+       "1015                         Studi 86\\nKaty ill             West Jue   \n",
+       "1043                              22 Kateen ova             Noth Joa   \n",
+       "1052                88, rue Stéphanie Letellier              Rouxnec   \n",
+       "1062                        Wenke-Hörle-Ring 36          Eggenfelden   \n",
+       "1120                    27 Rodgrs Rdgs Apt. 269       Suth Jeffererg   \n",
+       "1170  Calleón de Gregorio Bustamante 28 Piso 7               La Pala   \n",
+       "1339                                Hugasse 8/8             Hainichn   \n",
+       "1360                      chemin Denise Ferrand  Saint ChalotteVille   \n",
+       "1384                           753 Lindsey Pine          Thompsonshe   \n",
+       "1423                    91, rue Laetitia Collet     Dias-sur-Normand   \n",
+       "1433                             30 Mpy Ovepass           Jeiferview   \n",
+       "1452                           233 Kevin Street            Larryside   \n",
+       "1489                             Langernplatz 0        Stadtsteinach   \n",
+       "1544                    Detlef-Binner-Platz 0/1                 Burg   \n",
+       "1557               5163 Rebecca Creek Suite 421    North Theresaberg   \n",
+       "1567                        Vil e Imel Oorio 25                 Madd   \n",
+       "1616         Avnida d Grgorio Manón 344 Prta 8              Ponevedr   \n",
+       "1647             Pasadzo de Josep Bentez  Pso             Las Palmas   \n",
+       "1652                          HennyLorchRng 484       Hohensein-Ensh   \n",
+       "1703             721 Christopher View Suite 840          Watsonmouth   \n",
+       "1707                          85663 Savage Gles              Mcgeeon   \n",
+       "1722                  58513 Paricia Res Suie 45                So Me   \n",
+       "1759                    9490 Curts Spur Sute 82              Jseptwn   \n",
+       "1796                           57, enue de Gros             BenadBou   \n",
+       "1809                             Preißgasse 0/4                Soest   \n",
+       "1820                                    Eies. 7             Ba Lnwra   \n",
+       "1829                             Ladeckgasse 11         Rockenhausen   \n",
+       "1837                   258 Day Hollow Suite 410        Kimberlyhaven   \n",
+       "1914                     533 Fitzpatrick Bypass          Francesberg   \n",
+       "1938                      67, rue Inès Valentin              Nicolas   \n",
+       "1948                               Mitschkeee 8                   Mß   \n",
+       "1960           Calleón de Dolore Parea 21 At 7               Palncia   \n",
+       "1968                                   Stotr. 1              Saulgau   \n",
+       "1995           5361 Patterson Mission Suite 504           Villaburgh   \n",
+       "2010                  Christa-Ullrich-Allee 0/1      Schwäbisch Hall   \n",
+       "2075                    Laila-Scheibe-Allee 2/0          Luckenwalde   \n",
+       "\n",
+       "                    state                            email           user_name  \n",
+       "337                 Berli                 frncoise@wgnerco      christinefinke  \n",
+       "377             Thüringen                   inrharff@yah.d      walentinabeier  \n",
+       "506                    BO  halenewilliams@wilson-sandes.og               amy17  \n",
+       "545                Bayern               olav44@bolander.de           bettyhahn  \n",
+       "592             Thüringen      jaentschbirgitt@boerner.org         francesco44  \n",
+       "633                    AR               ilemaitre@voila.fr            astrid58  \n",
+       "658                 Texas      christina74@doyle-baker.biz               olynn  \n",
+       "723                Oregon         howardreginald@gmail.com              dana91  \n",
+       "762          South Dakota           nancyescobar@brown.net             ruben71  \n",
+       "772       Rheinland-Pfalz   auch-schlauchindietlind@gmx.de            zitakuhl  \n",
+       "783                    Iw                  jared52@aoo.com      thomasthompson  \n",
+       "889                   Lgo                       ev0@oo.com      colomerenrique  \n",
+       "928                    BS                   rdorm@dbmi.com              anne28  \n",
+       "979               La Rioa                   sldrpére@ps.cm           ipeñalver  \n",
+       "995              Colorado               jordan60@gmail.com         mendozajody  \n",
+       "1015                   VE                        em@mil.cm            bbennett  \n",
+       "1043                   KZ                asleig65@aisay.co            lloydann  \n",
+       "1052                   SE               boutineric@blin.fr             iwagner  \n",
+       "1062       Sachsen-Anhalt                 oluebs@troest.de              xklotz  \n",
+       "1120               Iinois             stepanie90@rogers.co              lori67  \n",
+       "1170            Salamanca      usolana@jáuregui-pedraza.om          gloriaolmo  \n",
+       "1339          Neersachsen                     jun@putz.com          jesselmaja  \n",
+       "1360                   IE               jchretien@costacom            joseph60  \n",
+       "1384              Colorao                  ashey28@rice.co              todd72  \n",
+       "1423                   SC        deschampsgabriel@guyot.fr        michelepetit  \n",
+       "1433                 Utah             insnashl@gas-hais.cm       christopher62  \n",
+       "1452              Indiana      gregg39@hernandez-gomez.com         knapprobert  \n",
+       "1489            Thüringen        haasemarieluise@noack.com              istoll  \n",
+       "1544        Niedersachsen              hanne-lore98@gmx.de            thomas14  \n",
+       "1557               Alaska      stephenwilliams@summers.com            ahawkins  \n",
+       "1567               Vicaya               mordóñ@cámara.info               eva16  \n",
+       "1616                 Lugo     icent4@montenero-brroso.info  sanmartínguillermo  \n",
+       "1647                  Mia                  srgio24@gail.co       luis-miguel23  \n",
+       "1652      BadenWürtteberg                  huerhes@hmal.de       anne-katrin51  \n",
+       "1703          Connecticut          chenjessica@hotmail.com          patricia34  \n",
+       "1707                  Als               bcaldern@htmail.cm         emilytorres  \n",
+       "1722               Alaska                      diuez@uess.               cmoss  \n",
+       "1759              Arizona                  ncole@yahoo.com              csmith  \n",
+       "1796                   AL                 rogrlro@munoz.om              xherve  \n",
+       "1809      Rheinland-Pfalz                 tloos@krause.net               abien  \n",
+       "1820              Thürige               rsthveriue@mies.rg           kwernecke  \n",
+       "1829  Nordrhein-Westfalen      grein-grotharnim@kallert.de           siegmar08  \n",
+       "1837             Colorado               brandy00@yahoo.com               amy30  \n",
+       "1914             Michigan        terrystephen@anderson.org               gcain  \n",
+       "1938                   FI               milletedith@sfr.fr            tthierry  \n",
+       "1948         SachsnAnhalt                  nrejwgner@gmx.e         muehlehenni  \n",
+       "1960             Cantaria                vázqzlornzo@al.om            daniel70  \n",
+       "1968          Shlsg-Holst            arlenpruschke@salz.or              kklemm  \n",
+       "1995         Rhode Island              kristen54@gmail.com               jkent  \n",
+       "2010               Hessen         gumprichalice@schmidt.de            fruppert  \n",
+       "2075              Hamburg     gutknechtevelyn@niemeier.com            dkreusel  "
+      ]
+     },
+     "execution_count": 13,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "customers[customers.duplicated([\"user_name\"])]\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "f268d35d-797f-49f3-9d24-63a83f62fc59",
+   "metadata": {},
+   "source": [
+    "Ein darauf den user_name gefilterter Datensatz kann so aussehen:\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "id": "46e76915-de2b-4227-aba6-0d53c43b651b",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>name</th>\n",
+       "      <th>job</th>\n",
+       "      <th>company</th>\n",
+       "      <th>street_address</th>\n",
+       "      <th>city</th>\n",
+       "      <th>state</th>\n",
+       "      <th>email</th>\n",
+       "      <th>user_name</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>236</th>\n",
+       "      <td>Aysel Binner</td>\n",
+       "      <td>Recycling officer</td>\n",
+       "      <td>Kuhl Kallert Stiftung &amp; Co. KGaA</td>\n",
+       "      <td>Beatrix-Kranz-Straße 5/4</td>\n",
+       "      <td>Fulda</td>\n",
+       "      <td>Berlin</td>\n",
+       "      <td>francoise22@wagner.com</td>\n",
+       "      <td>christinefinke</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>337</th>\n",
+       "      <td>Aysel Binner</td>\n",
+       "      <td>Reccig officer</td>\n",
+       "      <td>Kuhl Kalleww Swifwunw &amp; Co. KGaA</td>\n",
+       "      <td>Batix-Kanz-Staß 5/4</td>\n",
+       "      <td>Fulda</td>\n",
+       "      <td>Berli</td>\n",
+       "      <td>frncoise@wgnerco</td>\n",
+       "      <td>christinefinke</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "             name                job                           company  \\\n",
+       "236  Aysel Binner  Recycling officer  Kuhl Kallert Stiftung & Co. KGaA   \n",
+       "337  Aysel Binner     Reccig officer  Kuhl Kalleww Swifwunw & Co. KGaA   \n",
+       "\n",
+       "               street_address   city   state                   email  \\\n",
+       "236  Beatrix-Kranz-Straße 5/4  Fulda  Berlin  francoise22@wagner.com   \n",
+       "337       Batix-Kanz-Staß 5/4  Fulda   Berli        frncoise@wgnerco   \n",
+       "\n",
+       "          user_name  \n",
+       "236  christinefinke  \n",
+       "337  christinefinke  "
+      ]
+     },
+     "execution_count": 15,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "customers[customers[\"user_name\"] == \"christinefinke\"]\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "cd8ef668-a32a-41d7-840d-268baa03b2f7",
+   "metadata": {},
+   "source": [
+    "Nun können wir diejenigen Datenpunkte löschen, deren user_name identisch ist:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "id": "fe3c078d-e6cc-403d-9443-9415a798327d",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "(2080, 8)\n",
+      "(2029, 8)\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(customers.shape)\n",
+    "customers.drop_duplicates([\"user_name\"], inplace=True)\n",
+    "print(customers.shape)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "65a4323b-c93d-447a-82fd-45b0c4d85532",
+   "metadata": {},
+   "source": [
+    "Dies hat 2080-2029=51 Original-Datenpunkte gelöscht."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "78474f25-57ce-42e0-bed0-7f10fd6d61cc",
+   "metadata": {},
+   "source": [
+    "## 2.2. Deduplizieren mit fuzzywuzzy\n",
+    "\n",
+    "Als Alternative und zum Vergleich fürs Identifizieren von Duplikation und das entsprechende Deduplizieren schauen wir uns die Bibliothek fuzzywuzzy an: \n",
+    "\n",
+    "https://www.python4data.science/de/latest/clean-prep/string-matching.html\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "68d6b9d4-0fd7-4416-bca2-b6a2e42781ea",
+   "metadata": {},
+   "source": [
+    "\n",
+    " "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "31089ed7-9186-4c8d-86ff-5542cd444b1f",
+   "metadata": {},
+   "source": [
+    "## 2.3. Deduplizieren mit Dedupe\n",
+    "\n",
+    "Alternativ können wir die duplizierte Daten mit der Dedupe-Bibliothek (https://docs.dedupe.io/en/latest/) erkennen, die ein flaches neuronales Netzwerk verwendet, um aus einem kleinen Training zu lernen.\n",
+    "\n",
+    "Das entsprechende Notebook finden wir hier: \n",
+    "\n",
+    "https://www.python4data.science/de/latest/clean-prep/deduplicate.html#3.-Dedupe"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "41592878-3dac-4a9b-9929-1bd5c11f2de9",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/docs/2deep_ml_ops/index.rst b/docs/2deep_ml_ops/index.rst
index 8c96597..3d5e632 100644
--- a/docs/2deep_ml_ops/index.rst
+++ b/docs/2deep_ml_ops/index.rst
@@ -5,12 +5,51 @@ Vertiefung: Neuronale Netze, Deep Learning, Systemüberblick, Use Case Fallgrube
 
 - Zusammenfassung der Konzepte aus Tag 1: Überwachtes, unüberwachtes und bestärkendes Lernen.
 
-- Wichtige Begriffe: **Modell, Trainingsdaten, Testdaten, Overfitting, Regularisierung, Optimierung.**
+- Wichtige Begriffe: **Modell, Trainingsdaten, Testdaten.**
 
 - Diskussion der zentralen Herausforderungen im ML: **Bias, Datenqualität, Modellinterpretierbarkeit.**
 
 - Offene Fragen und Klarstellungen
 
+
+
+* Logistic Regression 
+* Decision Tree => Random Forest 
+* SVM ()
+
+.. list-table:: Vergleich von Regression, Klassifikation, Clustering und autonomen Aufgaben
+   :header-rows: 1
+
+   * - Merkmal
+     - :ref:`regression`
+     - :ref:`classification`
+     - :ref:`clustering`
+     - :ref:`autonomous`
+     - :ref:`generative`
+   * - Art der Vorhersage
+     - Kontinuierlicher Wert
+     - Diskrete Klassen
+     - Gruppenbildung ohne Labels
+     - Eigenständige Entscheidungsfindung
+     - Erzeugung neuer Daten
+   * - Use Case Beispiele
+     - Preisprognosen, Finanzanalysen
+     - Bilderkennung, Spam-Filter
+     - Kundensegmentierung, Anomalieerkennung
+     - Selbstfahrende Autos, Industrieroboter
+     - KI-generierte Bilder, Texte, Musik
+   * - Gängige "Klassischen" Algorithmen
+     - Lineare Regression, neuronale Netze
+     - Decision Trees (Entscheidungsbäume), Random Forest, SVM
+     - K-Means Clustering, K-nearest neighbour (kNN), DBSCAN
+     - Reinforcement Learning, Deep Learning
+     - GANs, VAEs, Transformer
+
+
+Heutiger Kursinhalt
+~~~~~~~~~~~~~~~~~~~~
+
+
 .. list-table:: 
    :header-rows: 1
 
@@ -18,6 +57,8 @@ Vertiefung: Neuronale Netze, Deep Learning, Systemüberblick, Use Case Fallgrube
      - Inhalte
    * - Recap der wichtigsten Themen
      - Wiederholung der ML-Konzepte, Diskussion, Quiz
+   * - Modellverhalten 
+     - Overfitting, Regularisierung und Optimierung
    * - Theorie: Neuronale Netze und Deep Learning
      - Aufbau, Backpropagation, Aktivierungsfunktionen
    * - Praxis: Bildklassifikation mit CNNs
@@ -31,7 +72,9 @@ Vertiefung: Neuronale Netze, Deep Learning, Systemüberblick, Use Case Fallgrube
    :titlesonly:
    :maxdepth: 0
 
+   deduplicate
+   modellverhalten
    neural-net
+   neural-net-optimize
    cnn-beispiel
    mlops
-   usecase-pitfalls
diff --git a/docs/2deep_ml_ops/modellverhalten.rst b/docs/2deep_ml_ops/modellverhalten.rst
new file mode 100644
index 0000000..c07d957
--- /dev/null
+++ b/docs/2deep_ml_ops/modellverhalten.rst
@@ -0,0 +1,243 @@
+Modellverhalten: Overfitting, Regularisierung und Optimierung
+================================================================
+
+Einführung
+---------------
+
+Nachdem wir uns am ersten Tag mit **Regression und Klassifikation** beschäftigt haben, wenden wir uns nun einem zentralen Problem im maschinellen Lernen zu: 
+**Wie stellen wir sicher, dass unser Modell nicht nur auf Trainingsdaten gut funktioniert, sondern auch auf unbekannte Daten?**
+
+Dazu müssen wir drei wichtige Konzepte verstehen:
+
+1. **Overfitting und Underfitting** – Wann ist ein Modell zu komplex oder zu einfach?
+2. **Regularisierung** – Wie können wir Overfitting verhindern?
+
+1. Overfitting und Underfitting
+----------------------------------
+
+Beim Training von ML-Modellen kann es vorkommen, dass ein Modell entweder **zu gut auf Trainingsdaten passt (Overfitting)** oder **nicht gut genug trainiert wird (Underfitting)**.
+
+- **Overfitting** tritt auf, wenn das Modell die Trainingsdaten **zu stark auswendig lernt** und nicht auf neue Daten generalisiert.
+- **Underfitting** tritt auf, wenn das Modell **zu simpel ist**, um die Muster in den Daten zu erfassen.
+
+
+Underfitting
+------------
+
+**Underfitting** tritt auf, wenn ein Modell zu einfach ist, um die zugrunde liegenden Muster in den Daten zu erfassen. 
+Dies geschieht oft, wenn das Modell zu wenige Parameter hat oder wenn es nicht ausreichend trainiert wurde. 
+Ein unterangepasstes Modell zeigt sowohl auf den Trainingsdaten als auch auf den Testdaten hohe Fehlerwerte, weil es die Struktur der Daten nicht richtig abbilden kann. 
+Ein typisches Beispiel ist die Verwendung einer linearen Regression für eine stark nicht-lineare Beziehung, wodurch das Modell die Zusammenhänge nicht ausreichend erfasst.
+
+
+Overfitting
+--------------
+
+Overfitting ist eines der größten Probleme im maschinellen Lernen. 
+
+Was ist Overfitting?
+::::::::::::::::::::::::
+
+Es tritt auf, wenn ein Modell die Trainingsdaten **zu stark auswendig lernt**, anstatt die zugrunde liegenden Muster zu erkennen. 
+Dies führt dazu, dass das Modell auf neuen, unbekannten Daten schlecht generalisiert.
+
+Das passiert leicht, wenn das ausgewählte und trainierte Modell zu komplex ist und sich **zu stark an den Trainingsdatensatz anpasst**, sodass es auf neuen Daten nicht mehr gut funktioniert. 
+
+**Beispiel:**
+
+Stellen wir uns vor, wir möchten den Zusammenhang zwischen der Körpergröße und dem Gewicht von Personen vorhersagen. 
+
+- Ein einfaches Modell (z. B. lineare Regression) könnte sagen: **„Größere Menschen wiegen tendenziell mehr.“**
+- Ein überangepasstes Modell könnte sich jedoch **zu stark an zufällige Ausreißer anpassen** und unnötig komplexe Regeln aufstellen wie: 
+
+**„Personen mit einer Körpergröße von 1,78m wiegen exakt 73,2kg, aber bei 1,79m sinkt das Gewicht plötzlich auf 71,5kg.“**
+
+Das überangepasste Modell wird zwar auf den Trainingsdaten eine sehr niedrige Fehlerquote haben, 
+**aber es wird auf neuen, unbekannten Daten sehr schlecht funktionieren**.
+
+.. figure:: ../_static/images/day2-overfitting.png
+   :alt: Illustration von Overfitting und Underfitting
+   :align: center
+   :width: 400px
+
+   **Abbildung 1:** Overfitting, citation: By Chabacano - Own work, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=3610704
+
+Eine gute Darstellung von Overfitting, Underfitting und einem gut generalisierten Modell kann `hier <https://de.mathworks.com/discovery/overfitting.html>`_ gefunden werden. 
+
+
+Wie erkennt man überangepasste Modelle?
+----------------------------------------
+
+**Overfitting** ist nicht immer sofort sichtbar, aber es gibt einige **Hinweise**, anhand derer man es erkennen kann:
+
+1. **Großer Unterschied zwischen Trainings- und Testfehler**
+   - Wenn der Fehler auf den Trainingsdaten sehr niedrig, aber auf den Testdaten hoch ist, liegt meist Overfitting vor.
+
+2. **Zu hohe Modellkomplexität**
+   - Wenn ein Modell sehr viele Parameter (Gewichte) hat, ist das Risiko hoch, dass es Overfitting betreibt.
+
+3. **Sehr gute Leistung auf bekannten, aber schlechte Leistung auf neuen Daten**
+   - Wenn das Modell sich perfekt an die Trainingsdaten anpasst, aber auf neuen Daten katastrophale Fehler macht.
+
+1. Trainings- und Testverlust:
+:::::::::::::::::::::::::::::::
+
+Definition
+~~~~~~~~~~~~
+
+**Training Loss (Trainingsverlust)**
+
+Der Trainingsverlust ist eine Metrik, mit der bewertet wird, wie gut ein Deep-Learning-Modell zu den Trainingsdaten passt. Das heißt, er bewertet den Fehler des Modells in der Trainingsmenge. Beachten Sie, dass der Trainingssatz ein Teil eines Datensatzes ist, der für das anfängliche Training des Modells verwendet wird. Rechnerisch wird der Trainingsverlust durch die Summe der Fehler für jedes Beispiel im Trainingssatz berechnet.
+
+Es ist auch wichtig zu beachten, dass der Trainingsverlust nach jedem Stapel gemessen wird. Dies wird in der Regel durch das Aufzeichnen einer Kurve des Trainingsverlustes veranschaulicht.
+
+
+**Validierungsverlust (Validation Loss)**
+
+Im Gegensatz dazu ist der Validierungsverlust eine Metrik, die verwendet wird, um die Leistung eines Deep-Learning-Modells auf dem Validierungsset zu bewerten. Der Validierungssatz ist ein Teil des Datensatzes, der beiseite gelegt wird, um die Leistung des Modells zu validieren. Der Validierungsverlust ist ähnlich wie der Trainingsverlust und wird aus der Summe der Fehler für jedes Beispiel im Validierungssatz berechnet.
+
+Zusätzlich wird der Validierungsverlust nach jeder Epoche gemessen. Dies gibt Aufschluss darüber, ob das Modell weiter optimiert oder angepasst werden muss oder nicht. Zu diesem Zweck wird normalerweise eine Lernkurve für den Validierungsverlust erstellt.
+
+Darstellung
+~~~~~~~~~~~~
+
+Eine typische Methode, um Overfitting zu erkennen, ist die **Lernkurve**, die zeigt, wie sich der Fehler für Trainings- und Testdaten über die Trainingszeit hinweg entwickelt:
+
+.. figure:: ../_static/images/day2-learning_curve.png
+   :alt: Lernkurve für Overfitting
+   :align: center
+   :width: 600px
+
+   **Abbildung 3:** Lernkurve – Overfitting zeigt sich durch eine große Lücke zwischen Training- und Testfehler.
+
+**Interpretation des Trainingsverlustes (Training Loss)** 
+
+Die Lernkurve eines Overfit-Modells weist zu Beginn einen sehr niedrigen Training-Loss auf, der mit dem Hinzufügen von Trainingsbeispielen allmählich leicht ansteigt und nicht abflacht.
+
+**Interpretieren des Validierungsverlustes (Validation Loss)**
+
+Die Lernkurve eines Overfit-Modells weist zu Beginn einen hohen Validierungsverlust auf, der mit dem Hinzufügen von Trainingsbeispielen allmählich sinkt und nicht abflacht. 
+Das deutet darauf hin, dass das Hinzufügen weiterer Trainingsbeispiele die Leistung des Modells bei ungesehenen Daten verbessern kann.
+Es ist auch zu erkennen, dass die Trainings- und Validierungsverluste weit voneinander entfernt sind, die sich bei Hinzufügen zusätzlicher Trainingsdaten einander annähern können
+
+
+2. Modelkomplexität
+::::::::::::::::::::
+
+Bei der Entscheidung der Modellkomplexität sind zwei Faktoren ausschlaggebend: 
+**Bias** vs. **Varianz** 
+
+**Bias**
+
+Sei Y der wahre Wert eines Parameters, und sei 
+Y hat sei ein Schätzer von Y auf der Grundlage einer Datenstichprobe. Dann ist die Verzerrung des Schätzers Y-hat ist gegeben durch:
+
+   .. math::
+      Bias(\hat{Y}) = E(\hat{Y}) - Y 
+
+wobei 
+
+.. math:: E(\hat{Y}) 
+    
+der Erwartungswert des Schätzers 
+
+.. math:: \hat{Y} 
+
+ist. Es ist die Messung des Modells, wie gut es zu den Daten passt. 
+
+**Geringe Verzerrung:**
+
+Eine geringe Verzerrung bedeutet, dass weniger Annahmen getroffen werden, um die Zielfunktion zu erstellen. 
+In diesem Fall stimmt das Modell gut mit dem Trainingsdatensatz überein.
+Hohe Verzerrung: Ein hoher Verzerrungswert bedeutet, dass mehr Annahmen zur Erstellung der Zielfunktion getroffen werden. 
+In diesem Fall stimmt das Modell nicht genau mit dem Trainingsdatensatz überein. 
+
+**Varianz**
+
+Die Varianz ist ein Streuungsmaß. Sie ist das Quadrat der Standardabweichung und stellt die Werte-Verteilung um den Mittelwert herum dar. 
+Berechnet wird die Varianz, indem die Summe der quadrierten Abweichungen aller Messwerte vom arithmetischen Mittel durch die Anzahl der Messwerte dividiert wird.
+
+   .. math::
+      Var(X) = \sigma^2 = \sum_{i=1}^{n} (x_i - \mu)^2 * p_i
+
+Das Zusammenspiel zwischen Bias und Varianz in Bezug auf Modelkomplexität kann wiefolgt dargestellt werden:
+
+.. figure:: ../_static/images/day2-model_complexity-bias_and_variance.png
+   :alt: Modelkomplexität - Bias vs Varianz
+   :align: center
+   :width: 500px
+
+   **Abbildung 1:** Modelkomplexität - Bias vs Varianz, citation: By Bigbossfarin - Own work, CC0, https://commons.wikimedia.org/w/index.php?curid=105307219
+
+
+
+2. Regularisierung: Verhinderung von Overfitting
+---------------------------------------------------------
+
+Methoden zur Regularisierung:
+:::::::::::::::::::::::::::::
+
+* **Mehr Trainingsdaten verwenden**
+   * Je mehr Daten ein Modell sieht, desto besser kann es generalisieren.
+   * Besonders wichtig bei Deep Learning.
+
+
+* **L1-Regularisierung (Lasso Regression)** 
+   Setzt einige Koeffizienten auf 0, um unwichtige Merkmale zu eliminieren.
+
+   .. math::
+      L_1 = \sum{|x_i|} 
+
+* **L2-Regularisierung (Ridge Regression)** 
+   Bestraft große Koeffizienten, um ein stabileres Modell zu erzeugen.
+
+   .. math::
+      L_1 = \sqrt{\sum{{x_i}^2}}
+
+
+* **Cross-Validation**
+   * K-Fold Cross Validation hilft, das Modell auf mehreren Trainings/Test-Splits zu überprüfen.
+
+
+* **Dropout (bei neuronalen Netzen)** 
+   Deaktiviert zufällig einige Neuronen während des Trainings.
+
+* **Early Stopping**
+   * Stoppt das Training, sobald sich die Testfehler nicht weiter verbessern.
+
+* **Datenaugmentierung** 
+     Erhöht die Datenvielfalt (besonders nützlich in Computer Vision).
+
+
+Code-Beispiel: Ridge & Lasso Regression
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. code-block:: python
+
+   from sklearn.linear_model import Ridge, Lasso
+   from sklearn.model_selection import train_test_split
+   from sklearn.metrics import mean_squared_error
+
+   # Trainings- und Testsplit
+   X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
+
+   # Ridge Regression
+   ridge = Ridge(alpha=1.0)
+   ridge.fit(X_train, y_train)
+   y_pred_ridge = ridge.predict(X_test)
+
+   # Lasso Regression
+   lasso = Lasso(alpha=0.1)
+   lasso.fit(X_train, y_train)
+   y_pred_lasso = lasso.predict(X_test)
+
+   # Fehlervergleich
+   print("MSE Ridge:", mean_squared_error(y_test, y_pred_ridge))
+   print("MSE Lasso:", mean_squared_error(y_test, y_pred_lasso))
+
+
+Fazit & Ausblick
+-----------------
+
+* Overfitting kann Modelle unbrauchbar machen – Regularisierung hilft.
+* Ridge und Lasso sind wichtige Werkzeuge für lineare Modelle.
diff --git a/docs/2deep_ml_ops/neural-net-optimize.rst b/docs/2deep_ml_ops/neural-net-optimize.rst
new file mode 100644
index 0000000..6da2281
--- /dev/null
+++ b/docs/2deep_ml_ops/neural-net-optimize.rst
@@ -0,0 +1,123 @@
+Neuronale Netze und Deep Learning – Optimierung
+============================================
+
+Optimierung: Modellparameter effizient anpassen
+---------------------------------------------------
+
+Die Wahl der richtigen **Optimierungsmethode** ist entscheidend für die Leistung eines Modells. Optimierung bedeutet, dass wir die **Modellparameter so anpassen, dass der Fehler minimiert wird** und das Modell möglichst gut generalisiert. Dabei gibt es verschiedene Ansätze, die in den folgenden Unterabschnitten erklärt werden.
+
+Grundlagen der Optimierung 
+:::::::::::::::::::::::::::::::::::::::::::::
+
+* Jedes Machine-Learning-Modell hat eine **Kosten- oder Verlustfunktion (Loss Function)**, die misst, wie gut oder schlecht das Modell Vorhersagen trifft. 
+* Ziel der Optimierung ist es, diese Verlustfunktion zu minimieren. 
+* Dabei suchen wir die optimalen Modellparameter (z. B. Gewichte in neuronalen Netzen oder Koeffizienten in linearen Modellen), sodass der Fehler möglichst klein ist.
+
+Gradient Descent – Das Fundament der Optimierung
+::::::::::::::::::::::::::::::::::::::::::::::::::
+
+**Gradient Descent (Gradientenabstieg)** ist eine der häufigsten Optimierungstechniken. 
+
+* Dessen Ziel liegt darin, den minimalen Verlustwert (Loss) zu finden. 
+* Dabei wird die **Ableitung der Verlustfunktion** verwendet, um herauszufinden, in welche Richtung sich die Parameter ändern müssen, um den Fehler zu minimieren.
+
+.. math::
+   w := w - \alpha \frac{\partial L}{\partial w}
+
+* **w**: Modellparameter (z. B. Gewichte in einem neuronalen Netz)
+* **L**: Verlustfunktion (Loss Function)
+* **\alpha**: Lernrate – bestimmt, wie große Schritte das Modell bei der Optimierung macht
+
+Varianten von Gradient Descent
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+1. **Batch Gradient Descent:**
+
+   * Verwendet **den gesamten Trainingsdatensatz**, um den Gradienten zu berechnen.
+   * **Vorteil:** Konvergiert stabil.
+   * **Nachteil:** Kann sehr langsam sein, insbesondere bei großen Datensätzen.
+
+2. **Stochastic Gradient Descent (SGD):**
+
+   * wird als **stochastische Approximation** der Gradientenabstiegsoptimierung betrachtet:
+   * sie ersetzt den tatsächlichen Gradient (berechnet aus dem gesamten Datensatz) durch eine Schätzung davon (berechnet aus einer **zufällig selektierten Teilmenge** der Daten) 
+   * **Vorteil:** Schneller als Batch Gradient Descent, da es nach jeder Instanz ein Update gibt.
+   * **Nachteil:** Kann aufgrund der ständigen Updates stark schwanken (stochastisches Verhalten).
+
+3. **Mini-Batch Gradient Descent:**
+
+   * Ein **Kompromiss zwischen Batch und SGD** – nutzt kleine **Gruppen (Batches)** von Daten.
+   * **Vorteil:** Stabiler als SGD, aber effizienter als Batch.
+   * **Nachteil:** Erfordert eine optimale Wahl der **Batch-Größe**.
+
+
+**Beispiel:** Auswirkungen der Lernrate auf das Training
+
+.. figure:: ../_static/images/day2-learning_rate_effect.png
+   :alt: Einfluss der Lernrate
+   :align: center
+   :width: 500px
+
+   **Abbildung 2:** Eine zu hohe Lernrate konvergiert nicht, eine zu niedrige dauert zu lange.
+
+Erweiterte Optimierungsverfahren**
+:::::::::::::::::::::::::::::::::::::::::::::
+
+Neben Gradient Descent gibt es weiterentwickelte Algorithmen, die schneller oder stabiler konvergieren:
+
+Momentum-based Methoden
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+**Momentum Optimizer:**
+
+* Verwendet einen **Impulswert**, um den Lernprozess zu beschleunigen und Schwankungen zu verringern.
+* Formell wird ein "Momentumeffekt" hinzugefügt, der die Aktualisierung der Gewichte glättet.
+
+.. math::
+   v_t = \beta v_{t-1} + (1-\beta) \nabla L(w)
+
+* **v_t**: Beschleunigung in Richtung des Minimums
+* **\beta**: Momentum-Faktor (meist um 0.9)
+
+Adaptiven Lernraten-Methoden
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+**Adagrad:**
+
+* Passt die Lernrate individuell für jede Modellparameter-Aktualisierung an.
+* **Vorteil:** Funktioniert gut für spärliche Daten.
+* **Nachteil:** Lernrate kann zu stark abfallen.
+
+**RMSprop:**
+
+* Eine Erweiterung von Adagrad mit **gleitendem Durchschnitt der Gradienten**.
+* Funktioniert besonders gut für **tiefe neuronale Netze**.
+
+**Adam-Optimizer (Adaptive Moment Estimation):**
+
+* Kombiniert die Vorteile von Momentum und RMSprop, d.h. passt eigenständig die Lernrate individuell für jede Modellparameter-Aktualisierung an, ohne dass die Lernrate zu stark abfällt. 
+* **Vorteil:** Stabile und schnelle Konvergenz, besonders in Deep Learning.
+* **Nachteil:** Kann für einige Probleme eine zu hohe Variabilität aufweisen.
+
+
+Hyperparameter-Tuning – Die richtige Balance finden
+------------------------------------------------------
+
+Neben der Optimierung der Modellparameter selbst müssen wir oft **Hyperparameter optimieren** (z. B. die Lernrate \( \alpha \)). Dazu gibt es verschiedene Methoden:
+
+**Grid Search:**
+
+* Testet systematisch alle möglichen Kombinationen von Hyperparametern.
+* **Nachteil:** Sehr rechenintensiv.
+
+**Random Search:**
+
+* Wählt zufällige Kombinationen von Hyperparametern aus.
+* **Vorteil:** Spart Rechenzeit, oft ähnlich gute Ergebnisse wie Grid Search.
+
+**Bayesian Optimization:**
+
+* Verwendet probabilistische Modelle zur effizienten Auswahl von Hyperparametern.
+* **Vorteil:** Schneller als Grid Search bei komplexen Modellen.
+
+
diff --git a/docs/2deep_ml_ops/neural-net.rst b/docs/2deep_ml_ops/neural-net.rst
index 8f35807..609d28c 100644
--- a/docs/2deep_ml_ops/neural-net.rst
+++ b/docs/2deep_ml_ops/neural-net.rst
@@ -1,17 +1,52 @@
 Neuronale Netze und Deep Learning – Theorie
-----------------------------------------------------------
+============================================
 
-**Grundlagen:**
+Grundlagen:
+----------------------------------------------------------
 
 Was sind künstliche neuronale Netze? Wie unterscheiden sie sich von klassischen ML-Modellen?
 
-- Künstliche neuronale Netze bestehen aus **Schichten von Neuronen**, die durch gewichtete Verbindungen miteinander verbunden sind.
+* Künstliche neuronale Netze bestehen aus **Schichten von Neuronen**, die durch gewichtete Verbindungen miteinander verbunden sind.
+* Aufbau eines neuronalen Netzes: **Eingabeschicht, versteckte Schichten, Aktivierungsfunktionen, Ausgangsschicht.**
+* Jedes Neuron führt eine Berechnung basierend auf einer **Aktivierungsfunktion** durch und gibt das Ergebnis an die nächste Schicht weiter.
+
+
+.. figure:: ../_static/images/day2-deepLearn.png
+   :alt: Illustration von Deep Neural Networks
+   :align: center
+   :width: 700px
+
+   **Abbildung 1:** Deep Neural Networks, source: `Lamarr-Institute <https://lamarr-institute.org/blog/deep-neural-networks/>`_, https://lamarr-institute.org/wp-content/uploads/deepLearn_2_EN.png 
+
+
+Wichtige Aktivierungsfunktionen:
+::::::::::::::::::::::::::::::::
+
+* **ReLU (Rectified Linear Unit):** 
+
+Häufig in CNNs verwendet, eliminiert negative Werte.
 
-- Aufbau eines neuronalen Netzes: **Eingabeschicht, versteckte Schichten, Aktivierungsfunktionen, Ausgangsschicht.**
+* **Sigmoid:** 
 
-- Jedes Neuron führt eine Berechnung basierend auf einer **Aktivierungsfunktion** durch und gibt das Ergebnis an die nächste Schicht weiter.
+Wandelt Werte in einen Bereich zwischen 0 und 1 um, nützlich für Wahrscheinlichkeitsprognosen.
 
-**Forwardpropagation (Vorwärtsdurchlauf):**
+* **Softmax:** 
+
+Wird in Klassifikationsproblemen für mehr als zwei Klassen genutzt.
+
+
+Fortgeschrittene Deep-Learning-Techniken:
+::::::::::::::::::::::::::::::::::::::::::
+
+* Convolutional Neural Networks (CNNs) für **Bildverarbeitung**.
+
+* Recurrent Neural Networks (RNNs) für **Sequenz- und Textverarbeitung**.
+
+* Transformer-Modelle für **NLP (z. B. BERT, GPT).**
+
+
+Forwardpropagation (Vorwärtsdurchlauf):
+----------------------------------------------------------
 
 1. Die Eingabedaten werden in das Netzwerk eingespeist.
 
@@ -27,83 +62,77 @@ Was sind künstliche neuronale Netze? Wie unterscheiden sie sich von klassischen
 5. Am Ende der Forward Propagation wird die Loss Function berechnet.
 
 
-**Loss Functions (Verlustfunktionen) und ihre Rolle:**
+Loss Functions (Verlustfunktionen) und ihre Rolle:
+----------------------------------------------------------
 
-- Eine **Loss Function** misst die Differenz zwischen der Vorhersage des Modells und dem tatsächlichen Wert.
+Eine **Loss Function** 
 
-- Sie gibt an, wie gut oder schlecht das Modell arbeitet.
+* misst die Differenz zwischen der Vorhersage des Modells und dem tatsächlichen Wert.
+* gibt an, wie gut oder schlecht das Modell arbeitet.
+* wird während der Backpropagation genutzt, um die Gewichte des Netzwerks zu aktualisieren.
 
-- Wird während der Backpropagation genutzt, um die Gewichte des Netzwerks zu aktualisieren.
+Typische Loss Functions:
+:::::::::::::::::::::::::
 
-**Typische Loss Functions:**
+Mean Squared Error (MSE)
+~~~~~~~~~~~~~~~~~~~~~~~~
 
-1. **Mean Squared Error (MSE)** – Wird für **Regressionsprobleme** verwendet:
+Wird für **Regressionsprobleme** verwendet:
 
    .. math::
       MSE = \frac{1}{n} \sum_{i=1}^{n} (y_i - \hat{y}_i)^2
 
-   - Bestraft größere Fehler überproportional.
+* Bestraft größere Fehler überproportional.
+
+* Gut für kontinuierliche Werte wie Preisprognosen.
 
-   - Gut für kontinuierliche Werte wie Preisprognosen.
+Cross-Entropy Loss
+~~~~~~~~~~~~~~~~~~~~~~
 
-2. **Cross-Entropy Loss** – Wird für **Klassifikationsprobleme** verwendet:
+Wird für **Klassifikationsprobleme** verwendet:
 
    .. math::
       L = -\sum y_i \log(\hat{y}_i)
 
-   - Erhöht die Strafe, wenn das Modell sehr sicher, aber falsch ist.
+* Erhöht die Strafe, wenn das Modell sehr sicher, aber falsch ist.
 
-   - Wird z. B. bei **Softmax-Klassifikationen** genutzt.
+* Wird z. B. bei **Softmax-Klassifikationen** genutzt.
 
-**Zusammenhang zwischen Loss Function und Gradient Descent:**
+Zusammenhang zwischen Loss Function und Gradient Descent:
+----------------------------------------------------------
 
-- Gradient Descent ist der Algorithmus, der die Gewichte des Modells so anpasst, dass die Loss Function minimiert wird.
+* Gradient Descent ist der Algorithmus, der die Gewichte des Modells so anpasst, dass die Loss Function minimiert wird.
 
-- Die Ableitung der Loss Function bestimmt die Richtung, in die die Gewichte aktualisiert werden.
+* Die Ableitung der Loss Function bestimmt die Richtung, in die die Gewichte aktualisiert werden.
 
-- **Formel für das Gewicht-Update:**
+Formel für das Gewicht-Update:
+:::::::::::::::::::::::::::::::
+
+Das Wichtigste an einem Modell sind die **Gewichte (weights)**. 
+Diese werden während der **Backwardpropagation (Rückwärtsdurchlauf)**-Phase aktualisiert, 
+indem man das Produkt der Learning-Rate (\alpha) und der Ableitung der Loss-Funktion 
+von den aktuellen Gewichten subtrahiert: 
 
   .. math::
      w := w - \alpha \frac{\partial L}{\partial w}
 
-  wobei \( \alpha \) die Lernrate ist.
-
-- Dies passiert während der **Backwardpropagation (Rückwärtsdurchlauf):**
 
-**Backwardpropagation (Rückwärtsdurchlauf):**
 
-1. Der Fehler des Netzwerks wird berechnet, indem die Differenz zwischen der vorhergesagten und der tatsächlichen Ausgabe bestimmt wird, d.h. indem die Loss Function evaluiert wird.
+Backwardpropagation (Rückwärtsdurchlauf):
+----------------------------------------------------------
 
-2. Die Fehler werden von der letzten Schicht zurück durch das Netzwerk propagiert, um **die Gewichte der Neuronen zu aktualisieren**.
+* Nachdem in der Forwardpropagation die Input-Daten durch das Neural Network propagiert sind, werden die Ausgaben des Networks mit den gewünschten Ausgaben verglichen. 
+* Deren Differenz wird als der Fehler des Netzwerks in diesem Durchgang erachtet, d.h. Loss. 
+* Dieser Fehler wird dann in der `Backwardpropagation <https://de.wikipedia.org/wiki/Backpropagation>`_ über die letzten Schicht (Ausgabeschicht) zurück zur Eingabeschicht durch das Netzwerk propagiert, um **die Gewichte der Neuronen zu aktualisieren**.
 
-3. Die Berechnung erfolgt mit Hilfe der **Kettenregel der Ableitungen**, um die Gradienten für jedes Gewicht zu bestimmen:
+* Die Berechnung erfolgt mit Hilfe der **Kettenregel der Ableitungen**, um die Gradienten für jedes Gewicht zu bestimmen:
 
 ..   .. math::
 ..      rac{\partial L}{\partial w} = rac{\partial L}{\partial y} \cdot rac{\partial y}{\partial z} \cdot rac{\partial z}{\partial w}
 
-4. Durch die Anwendung des **Gradientenabstiegsverfahrens (Gradient Descent)** werden die Gewichte so angepasst, dass der Gesamtfehler des Netzwerks iterativ minimiert und somit das Modell optimiert wird.
-
-**Optimierungstechniken:**
-
-- **Stochastischer Gradientenabstieg (SGD):** Berechnet Gradienten basierend auf einer zufälligen Stichprobe aus den Trainingsdaten.
-
-- **Adam-Optimizer:** Eine erweiterte Form des SGD, die adaptives Lernen ermöglicht und schneller konvergiert.
-
-- **Momentum-Methoden:** Nutzen vergangene Gradienteninformationen, um stabilere Updates durchzuführen.
-
-**Wichtige Aktivierungsfunktionen:**
-
-- **ReLU (Rectified Linear Unit):** Häufig in CNNs verwendet, eliminiert negative Werte.
-
-- **Sigmoid:** Wandelt Werte in einen Bereich zwischen 0 und 1 um, nützlich für Wahrscheinlichkeitsprognosen.
-
-- **Softmax:** Wird in Klassifikationsproblemen für mehr als zwei Klassen genutzt.
-
-
-**Fortgeschrittene Deep-Learning-Techniken:**
-
-- Convolutional Neural Networks (CNNs) für **Bildverarbeitung**.
+* In jeder Epoche passt das Modell diese Parameter an und verringert so den Verlust, indem es dem Fehlergradienten folgt. 
+`Backpropagation <https://www.geeksforgeeks.org/backpropagation-in-neural-network/>`_ verwendet häufig Optimierungsalgorithmen wie Gradientenabstieg (Gradient Descent) oder stochastischer Gradientenabstieg (stochastic Gradient Descent). 
+* Der Algorithmus berechnet den Gradienten mit Hilfe der Kettenregel aus der Infinitesimalrechnung und kann so effektiv durch komplexe Schichten im neuronalen Netz navigieren, um die Kostenfunktion zu minimieren.
 
-- Recurrent Neural Networks (RNNs) für **Sequenz- und Textverarbeitung**.
+* Somit werden während der Backpropagation-Phase durch die Anwendung von z.B. **Gradientenabstiegsverfahrens (Gradient Descent)**  die Gewichte so angepasst, dass der Gesamtfehler des Netzwerks iterativ minimiert und somit das Modell optimiert wird.
 
-- Transformer-Modelle für **NLP (z. B. BERT, GPT).**
diff --git a/docs/4gen_ai/cusy_genai.ipynb b/docs/4gen_ai/cusy_genai.ipynb
deleted file mode 100644
index c55a287..0000000
--- a/docs/4gen_ai/cusy_genai.ipynb
+++ /dev/null
@@ -1,10 +0,0 @@
-{
- "cells": [],
- "metadata": {
-  "language_info": {
-   "name": "python"
-  }
- },
- "nbformat": 4,
- "nbformat_minor": 2
-}
diff --git a/docs/_static/images/day2-deepLearn.png b/docs/_static/images/day2-deepLearn.png
new file mode 100644
index 0000000..9b530c0
Binary files /dev/null and b/docs/_static/images/day2-deepLearn.png differ
diff --git a/docs/_static/images/day2-learning_curve.png b/docs/_static/images/day2-learning_curve.png
new file mode 100644
index 0000000..54a257f
Binary files /dev/null and b/docs/_static/images/day2-learning_curve.png differ
diff --git a/docs/_static/images/day2-learning_rate_effect.png b/docs/_static/images/day2-learning_rate_effect.png
new file mode 100644
index 0000000..e031476
Binary files /dev/null and b/docs/_static/images/day2-learning_rate_effect.png differ
diff --git a/docs/_static/images/day2-model_complexity-bias_and_variance.png b/docs/_static/images/day2-model_complexity-bias_and_variance.png
new file mode 100644
index 0000000..b9c9b45
Binary files /dev/null and b/docs/_static/images/day2-model_complexity-bias_and_variance.png differ
diff --git a/docs/_static/images/day2-overfitting.png b/docs/_static/images/day2-overfitting.png
new file mode 100644
index 0000000..66850b5
Binary files /dev/null and b/docs/_static/images/day2-overfitting.png differ