https://github.com/selfint/neat_openaigym
Using NEAT to solve OpenAI gym environments
https://github.com/selfint/neat_openaigym
neat-python neuroevolution openai-gym python3
Last synced: 3 months ago
JSON representation
Using NEAT to solve OpenAI gym environments
- Host: GitHub
- URL: https://github.com/selfint/neat_openaigym
- Owner: selfint
- Created: 2019-09-08T19:02:03.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2019-09-16T17:33:22.000Z (almost 6 years ago)
- Last Synced: 2025-02-14T08:49:12.327Z (4 months ago)
- Topics: neat-python, neuroevolution, openai-gym, python3
- Language: Python
- Size: 35.2 KB
- Stars: 2
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Objective
Implement the NEAT algorithm as elegantly as possible, and use it on
OpenAI-gym environments.
Based on this paper: .
# Algorithm Outline
## Testing the agents
1. Each agent is represented as a genome, which contains the nodes and connections
of that genome.
2. The simulation will generate an agent using each genome in its population list, let
it run in the environment (in turn) and return its accumulated reward.
## Generating a new generation
1. Adjust each genomes fitness based on its species (formula 2 page 13).
2. Generate new genomes using the rules specified in pages 11 - 13.
3. Split the genomes into species based on their 'genetic distance' (formula 1 page 13).