PyTorch NEAT

A PyTorch implementation of Kenneth O. Stanley's Neuroevolution of Augmenting Topologies (NEAT) paper.

Why

This past summer (2018) I dived into machine learning research and have undertaken an independent research project into catastrophic forgetting in neural networks with two partners. Part of our initial strategy requires evolving networks which requires an algorithm like NEAT. As a result of this implementation I am now very comfortable with PyTorch and have a near complete understanding of NEAT.

Code

Of Course! Go ahead and pick it apart on github.

Tools

Python, PyTorch

Example XOR Experiment

In the gif below, the foundational XOR problem is solved by this NEAT implmentation.