Better exploration with parameter noise
OpenAI Blog
Researchers added adaptive noise to reinforcement learning algorithm parameters and found it improved performance across tests. The technique showed consistent gains with minimal implementation complexity and no documented cases of performance degradation. This suggests parameter noise should become a standard exploration method to test on new reinforcement learning problems.
Why it matters
We’ve found that adding adaptive noise to the parameters of reinforcement learning algorithms frequently boosts performance. This exploration method is simple to implement and very rarely decreases performance, so it’s worth trying on any problem.