A Reinforcement learning based NEAT approach on OpenAI Gym's Car Racing game

Each folder contains one experiment:

• baseline considers a population of 32 individuals, FeedForward NN, speciation threshold of 3.0, elitism of 2, and 0 units in the initial hidden layer
• Elitism: baseline with elitism = 16
• Hidden 20 units: baseline with hidden layer initialized with 20 units
• Recurrent: baseline allowing Recurent neural networks
• Speciation: baseline with specieation threshold of 1.0
• Population 100: baseline with population of 100 individuals
• Population 100 RNN: Population 100 allowing Recurent neural networks
• Population 100 Speciation: Population 100 with speciation threshold set to 1.0
• Population 100 RNN Speciation: Population 100 with speciation threshold set to 1.0 and allowing recurrent neural networks.

A requirements file is placed inside each folder

To execute any experiment, install the requirements with pip install -r requirements.txt and run the code with: python name_of_file.py (for example to run baseline: python baseline.py)