MNIST Neural Network Model

This project demonstrates the creation, training, and evaluation of a neural network using TensorFlow and the MNIST dataset. The MNIST dataset is a collection of 28x28 grayscale images of handwritten digits (0-9) with corresponding labels.

Features

• Loading and preprocessing the MNIST dataset.
• Building a neural network with:
  • An input layer.
  • Two hidden layers (128 and 64 neurons, respectively).
  • An output layer with 10 neurons (one for each digit class).
• Training the model using the Adam optimizer and categorical cross-entropy loss.
• Evaluating the model's performance on a test dataset.
• Visualizing the training and validation accuracy/loss.
• Saving and loading the trained model.
• Making predictions on test data.

Files

• Tensorflow_Participation_Assignment_Assignment.ipynb: The main notebook containing the code for loading data, building, training, and evaluating the neural network.

Requirements

The project requires the following Python libraries:

• TensorFlow
• NumPy
• Matplotlib
• Pandas

To install the required libraries, run:

pip install tensorflow numpy matplotlib pandas

Usage

1. Clone this repository to your local machine.
2. Open the mnist_model.ipynb notebook in Jupyter Notebook or any compatible editor.
3. Run all cells in the notebook to:
   • Load and preprocess the MNIST dataset.
   • Build and train the model.
   • Save and load the model.
   • Evaluate and visualize the model's performance.

Model Architecture

• Input Layer: Flattened input of size 784 (28x28).
• Hidden Layer 1: Dense layer with 128 neurons and ReLU activation.
• Hidden Layer 2: Dense layer with 64 neurons and ReLU activation.
• Output Layer: Dense layer with 10 neurons and Softmax activation.

Results

The model achieves a high accuracy on the MNIST test dataset, showcasing its ability to classify handwritten digits effectively. Detailed training and validation metrics are visualized in the notebook.

Visualization

The notebook includes plots for:

• Training and validation accuracy over epochs.
• Training and validation loss over epochs.

License

This project is open-source and available under the MIT License.

Author

Developed by Samuel Dushime and Willy Kalisa

Commits

For any questions or suggestions, please feel free to reach out!