Merge pull request #1090 from NA-V10/improve-lenet-docs

Add documentation and comments for LeNetMNIST CNN example

Author: agibsonccc

dl4j-examples/src/main/java/org/deeplearning4j/examples/quickstart/modeling/convolution/LeNetMNIST.java

@@ -17,6 +17,12 @@
  * SPDX-License-Identifier: Apache-2.0
  ******************************************************************************/
 
+ // LeNetMNIST.java
+ // Classic LeNet-style CNN for MNIST digit classification.
+ // Expected accuracy: 98–99%.
+ // MNIST images: 28x28 grayscale (1 channel).
+
+
 package org.deeplearning4j.examples.quickstart.modeling.convolution;
 
 import org.apache.commons.io.FilenameUtils;
@@ -64,6 +70,10 @@ public static void main(String[] args) throws Exception {
          */
         log.info("Build model....");
 
+        // Build LeNet CNN architecture:
+       // Conv → Pool → Conv → Pool → Dense → Output
+
+
         MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder()
                 .seed(seed)
                 .l2(0.0005)
@@ -116,7 +126,9 @@ row vector format (i.e., 1x784 vectors), hence the "convolutionalFlat" input typ
 
         MultiLayerNetwork model = new MultiLayerNetwork(conf);
         model.init();
-
+        
+        // Evaluate the model at the end of each epoch using EvaluativeListener
+        
         log.info("Train model...");
         model.setListeners(new ScoreIterationListener(10), new EvaluativeListener(mnistTest, 1, InvocationType.EPOCH_END)); //Print score every 10 iterations and evaluate on test set every epoch
         model.fit(mnistTrain, nEpochs);

dl4j-examples/src/main/java/org/deeplearning4j/examples/quickstart/modeling/convolution/LeNetMNISTReLu.java

@@ -17,6 +17,10 @@
  * SPDX-License-Identifier: Apache-2.0
  ******************************************************************************/
 
+   // LeNetMNISTReLu.java
+  // Same as LeNetMNIST but uses ReLU activation instead of Tanh.
+
+
 package org.deeplearning4j.examples.quickstart.modeling.convolution;
 
 import org.datavec.api.io.labels.ParentPathLabelGenerator;

dl4j-examples/src/main/java/org/deeplearning4j/examples/quickstart/modeling/convolution/README.md

@@ -0,0 +1,57 @@
+# Convolutional Neural Network (CNN) Examples – DeepLearning4J
+
+This folder contains convolutional neural network (CNN) examples implemented using DeepLearning4J.
+
+## 🧠 LeNetMNIST.java
+Trains a LeNet-style CNN on the MNIST handwritten digit dataset.
+
+### ✔ What this example demonstrates
+- Loading the MNIST dataset
+- Building a classic LeNet CNN architecture
+- Training the network
+- Evaluating accuracy
+
+### ✔ Expected Accuracy
+**98%–99%** after 1–2 epochs.
+
+---
+
+## 📦 How to Run
+
+Build the project:
+
+mvn clean package
+
+Run the example:
+
+mvn -q exec:java -Dexec.mainClass="org.deeplearning4j.examples.quickstart.modeling.convolution.LeNetMNIST"
+
+Run the ReLU variant:
+
+mvn -q exec:java -Dexec.mainClass="org.deeplearning4j.examples.quickstart.modeling.convolution.LeNetMNISTReLu"
+
+---
+
+## 🧱 LeNet Architecture Breakdown
+- Convolution layer (20 filters)
+- Subsampling (2×2)
+- Convolution layer (50 filters)
+- Subsampling
+- Dense layer
+- Output layer (Softmax, 10 classes)
+
+---
+
+## 📂 Other Files
+| File | Description |
+|------|-------------|
+| **LeNetMNISTReLu.java** | LeNet variant with ReLU activation |
+| **CenterLossLeNetMNIST.java** | LeNet with center loss |
+| **CIFARClassifier.java** | CIFAR-10 image classifier |
+| **Conv1DUCISequenceClassifier.java** | 1D CNN example for sequences |
+
+---
+
+## 🙌 Why This Documentation Helps
+These CNN examples previously had no explanation, run instructions, or architecture summary.  
+This README improves clarity for new users and first-time contributors.