N-Layer Neural Network Framework

A feed-forward neural network framework written from scratch in C++, with no machine learning library underneath it. The architecture is configurable to any number of hidden layers, activation functions swap between sigmoid, tanh and linear, and training runs on backpropagation with gradient descent. Networks are defined in external config files and weights save and load, so a trained model outlives the process that made it. I used it for handwritten digit classification.