Update binary_logistic_regression_tf.ipynb

Author: fs446

binary_logistic_regression_tf.ipynb

@@ -45,7 +45,7 @@
     "from sklearn.datasets import make_classification\n",
     "from sklearn.model_selection import train_test_split\n",
     "import tensorflow as tf\n",
-    "import tensorflow.keras as keras\n",
+    "from tensorflow import keras\n",
     "\n",
     "\n",
     "print(\n",
@@ -98,12 +98,23 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "e99a940f",
+   "id": "c795c19c-aa94-48b9-9385-390fd831a1fa",
    "metadata": {},
    "outputs": [],
    "source": [
-    "def predict_class(y):\n",
-    "    y[y < 0.5], y[y >= 0.5] = 0, 1"
+    "def predict_class_tf(y):\n",
+    "    y[y[:, 0] < 0.5, :], y[y[:, 0] >= 0.5, :] = 0, 1"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "8388f542-910c-408f-9df7-b9ef4ed6f02a",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def predict_class_my(y):\n",
+    "    y[:, y[0, :] < 0.5], y[:, y[0, :] >= 0.5] = 0, 1"
    ]
   },
   {
@@ -116,7 +127,7 @@
     "def evaluate(y_true, y_pred):\n",
     "    y_true_tmp = np.copy(y_true)\n",
     "    y_pred_tmp = np.copy(y_pred)\n",
-    "    predict_class(y_pred_tmp)\n",
+    "    predict_class_my(y_pred_tmp)\n",
     "\n",
     "    # https://www.tensorflow.org/api_docs/python/tf/math/confusion_matrix\n",
     "    # The matrix columns represent the prediction labels.\n",
@@ -193,7 +204,7 @@
     ")\n",
     "X_train, X_test, Y_train, Y_test = train_test_split(\n",
     "    X, Y, train_size=train_size, random_state=None\n",
-    ")"
+    ")\n"
    ]
   },
   {
@@ -221,6 +232,26 @@
     "print(\"X train dim\", X_train_our.shape, \"Y train dim\", Y_train_our.shape)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "a3aedc53-87b0-4e3c-91a7-73be1df13b95",
+   "metadata": {},
+   "source": [
+    "- prep for TF"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "3d4f778d-786a-4146-a662-4bbafa8577e0",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "Y_train = Y_train[:, None]  # newer TF needs  (x,1) instead of (x) arrays\n",
+    "Y_test = Y_test[:, None]\n",
+    "X_train.shape, X_test.shape, Y_train.shape, Y_test.shape"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "f2b9c6e6",
@@ -439,13 +470,17 @@
    "source": [
     "# prediction after training finished\n",
     "Y_train_pred_tf = model.predict(X_train)\n",
-    "predict_class(Y_train_pred_tf)\n",
+    "predict_class_tf(Y_train_pred_tf)\n",
+    "\n",
+    "print(Y_train_pred_tf.shape, Y_train.shape)\n",
     "\n",
     "# confusion matrix\n",
     "cm_train_tf = tf.math.confusion_matrix(\n",
-    "    labels=Y_train, predictions=Y_train_pred_tf, num_classes=2\n",
+    "    labels=np.squeeze(Y_train), predictions=np.squeeze(Y_train_pred_tf), num_classes=2\n",
     ")\n",
     "\n",
+    "\n",
+    "\n",
     "# get technical measures for the trained model on the training data set\n",
     "results_train_tf = model.evaluate(\n",
     "    X_train, Y_train, batch_size=M_train, verbose=verbose\n",
@@ -552,7 +587,7 @@
     "print(\"\\nm_test\", M_test)\n",
     "# our implementation needs transposed data\n",
     "X_test_our = X_test.T\n",
-    "Y_test_our = Y_test[None, :]\n",
+    "Y_test_our = Y_test.T\n",
     "print(\"X test dim\", X_test_our.shape, \"Y test dim\", Y_test_our.shape)"
    ]
   },
@@ -601,11 +636,11 @@
    "source": [
     "# prediction\n",
     "Y_test_pred_tf = model.predict(X_test)\n",
-    "predict_class(Y_test_pred_tf)\n",
+    "predict_class_tf(Y_test_pred_tf)\n",
     "\n",
     "# confusion matrix\n",
     "cm_test_tf = tf.math.confusion_matrix(\n",
-    "    labels=Y_test, predictions=Y_test_pred_tf, num_classes=2\n",
+    "    labels=np.squeeze(Y_test), predictions=np.squeeze(Y_test_pred_tf), num_classes=2\n",
     ")\n",
     "\n",
     "# get technical measures for the trained model on the training data set\n",
@@ -668,6 +703,16 @@
     "print(\"TF confusion matrix in %\\n\", cm_test_tf / M_test * 100.0)"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "e37d308d-9be7-4394-a7fd-d08a1feb279b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "X_train.shape, Y_train.shape, X_test.shape, Y_test.shape"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -684,8 +729,8 @@
     "\n",
     "    plt.figure(figsize=(10, 10))\n",
     "    plt.subplot(2, 1, 1)\n",
-    "    plt.plot(X_train[Y_train == 0, 0], X_train[Y_train == 0, 1], \"C0o\", ms=1)\n",
-    "    plt.plot(X_train[Y_train == 1, 0], X_train[Y_train == 1, 1], \"C1o\", ms=1)\n",
+    "    plt.plot(X_train[Y_train[:, 0] == 0, 0], X_train[Y_train[:, 0] == 0, 1], \"C0o\", ms=1)\n",
+    "    plt.plot(X_train[Y_train[:, 0] == 1, 0], X_train[Y_train[:, 0] == 1, 1], \"C1o\", ms=1)\n",
     "    plt.contourf(f1, f2, tmp, cmap=\"RdBu_r\")\n",
     "    plt.axis(\"equal\")\n",
     "    plt.colorbar()\n",
@@ -694,8 +739,8 @@
     "    plt.ylabel(\"feature 2\")\n",
     "\n",
     "    plt.subplot(2, 1, 2)\n",
-    "    plt.plot(X_test[Y_test == 0, 0], X_test[Y_test == 0, 1], \"C0o\", ms=1)\n",
-    "    plt.plot(X_test[Y_test == 1, 0], X_test[Y_test == 1, 1], \"C1o\", ms=1)\n",
+    "    plt.plot(X_test[Y_test[:, 0] == 0, 0], X_test[Y_test[:, 0] == 0, 1], \"C0o\", ms=1)\n",
+    "    plt.plot(X_test[Y_test[:, 0] == 1, 0], X_test[Y_test[:, 0] == 1, 1], \"C1o\", ms=1)\n",
     "    plt.contourf(f1, f2, tmp, cmap=\"RdBu_r\")\n",
     "    plt.axis(\"equal\")\n",
     "    plt.colorbar()\n",
@@ -735,7 +780,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.10.6"
+   "version": "3.12.3"
   }
  },
  "nbformat": 4,