Handwritten Digit Classifier

Deep neural network written in Python with NumPy and trained on the MNIST dataset: http://yann.lecun.com/exdb/mnist/.

Description

• Architecture: 3 hidden layers with 500, 300, and 100 neurons respectively
• Activation function: ReLU for hidden layers, softmax for output layer
• Optimization: Adam optimization algorithm, mini-batch size = 128
• Regularization: Inverted Dropout (keep_prob = 0.8)
• Features: Gradient checking

Built With

• Python
• NumPy

Authors

• Matthias Wright

License

This project is licensed under the MIT License - see the LICENSE.md file for details