This is a Jupyter Notebook that walks through creating a Neural Network from scratch, only using NumPy. It covers many of the basic concepts such as Neural Network structure, forward propagation, back propagation, activation functions, and learning rates. It builds a model that classifies the MNIST handwritten digit data set with ~90% accuracy.
All code is provided under the BSD 3-Clause license.
This project is maintained by @hodgesmr.
Please use it for good, not evil.