NEAT Tetris Agent (Unity / C#)
An application of NEAT (NeuroEvolution of Augmenting Topologies) to learning Tetris game policies: rather than fixing a network architecture and training its weights, NEAT evolves the topology and the weights together, growing complexity only when it pays off.
Multiple agent variants were implemented and compared in a Unity/C# environment, continuing a line of work that began with my BSc thesis on training intelligent agents with neuroevolution.
Stack: Unity, C#, NEAT
