§ feed · storyline
Evolution strategies as a scalable alternative to reinforcement learning
OpenAI finds evolution strategies match standard reinforcement learning performance on Atari and MuJoCo benchmarks while avoiding several common RL drawbacks.
We’ve discovered that evolution strategies (ES), an optimization technique that’s been known for decades, rivals the performance of standard reinforcement learning (RL) techniques on modern RL benchmarks (e.g.
Atari/MuJoCo), while overcoming many of RL’s inconveniences.
§ sources1 publication · timeline below