Overview

This corner describes how to build a neural network library in Javascript. Given that it is implemented in Javascript, it will be fairly easy to work with the API, however, it will be really slow compared to the libraries written in optimized C or ones that utilize a GPU.

The below implementation is implemented in the backpropagation library.

Topics

• Layer Implementation - description of how to implement the layer architecture of a neural network.
  • Regularization
• Network Implementation - shows how to integrate a list of layers into a neural network.
• Error Implementation - shows how implement a loss function.
• Backpropagation - shows the basic implementation of the backpropagation algorithm.
• Testing - tests the results on a simple example.