Genetic Algorithm Induction of Neural Networks (GAINN) for Maze Traversal

In this project we seek to automatically induce algorithms to navigate an agent through a nondeterministic maze-type environment. The method uses a genetic algorithm to construct and optimize a neural network controller for the agent. Hybridizing a genetic approach with a neural network has advantages because the genetic search for solutions can then integrate experience or memory into the decision making that would otherwise be very difficult with pure genetic algorithms.