A neural network trained to identify handwritten numbers 0-9

Developed with two classmates, Angelo and Winry, as part of a class project. We learned all about neural networks:

• Perceptron vs. Sigmoid Neurons
• Backpropogation
• Gradient Descent
• Alternative activation functions
• Relu
• Tanh
• Softmax
• Undefitting vs. Overfitting
• L1 and L2 regularization
• Kfolding
• Dropout

Trained using the MNIST dataset:

Our best training results had an accuracy of 97.11%