Built Neural Network to recognize handwritten digits
Built neural network in python from scratch using pandas and numpy to recognize handwritten digits using the MNIST dataset in CSV form. Used mini-batch gradient descent as optmization algorithm and used a 3 layer network with layer sizes of 784, 270, 10, respectively. Mini-batch size was 64 and 10 epochs were used with an average accuracy score of 90%. Implementation was fully vectorized, with a sigmoid activation function and softmax function for output layer. Object-Oriented program that allows for changing and experimenting with hyperparameters more easily.