Maze Solver
Version .94
This is a python implementation of a neural network system based on the Neural Evolution of Adaptive Topologies (NEAT) algorithm. It contains all the code to breed and evolve neural nets to solve 2-d mazes and a wrapper to demonstrate one of the older neural networks running through randomly generates mazes.
12/09/2006 - 224KB | Download
Requirements: Python must be installed and there must be a path to \Python\bin for the MazeSolver.bat file to run properly.
Usage: Start the demonstration with MazeSolver.bat
The neural network included in the package works best with mazes around 20 by 20 and 200 to 300 steps.
This demonstration neural network was build before some code refinements and is outdated and preforms poorly. I will create a more refined default neural network when I get some extra time.
