openai/evolution-strategies-starter
A distributed implementation of Evolution Strategies, a population-based optimization algorithm for training neural networks as an alternative to reinforcement learning.
Not currently ranked — collecting fresh signals.
star history
This repository contains code for the paper ‘Evolution Strategies as a Scalable Alternative to Reinforcement Learning.’ It implements a distributed master-worker architecture where the master broadcasts parameters to workers and receives returns. The implementation is designed to run on EC2 using AWS spot instances for cost-effective distributed training, supporting experiments like humanoid scaling as described in the original paper.
Frequently asked
- What is openai/evolution-strategies-starter?
- A distributed implementation of Evolution Strategies, a population-based optimization algorithm for training neural networks as an alternative to reinforcement learning.
- Is evolution-strategies-starter open source?
- Yes — openai/evolution-strategies-starter is open source, released under the MIT license.
- What language is evolution-strategies-starter written in?
- openai/evolution-strategies-starter is primarily written in Python.
- How popular is evolution-strategies-starter?
- openai/evolution-strategies-starter has 1.6k stars on GitHub.
- Where can I find evolution-strategies-starter?
- openai/evolution-strategies-starter is on GitHub at https://github.com/openai/evolution-strategies-starter.