marwanh7 / neural_network_numpy

In this repository I create a two layer neural network using numpy and train it on the MNIST digit recognizer data set.

The goal of this repo is to:

• Derive the equations for each component of the neural netowrk. (Forward propagation, backward propagation, parameter updates)
• Implement these equation to get a better understanding of the math and how neural networks operate under the hood
• Use this numpy neural network to predict on the MNIST digit data set.
• Use this as a starting point to create more complex neural networks using Tensorflow and Pytorch to solve interesting real world problems. Mainly focused on time series forcasting LSTM and computer vision CNN