OpenAI Reinforcement Learning Tools

5 articles • Started over 10 years ago • Last updated almost 7 years ago

OpenAI released a suite of reinforcement learning tools and benchmarks between 2016 and 2019 to advance AI research. The company launched OpenAI Gym in April 2016 as a toolkit for developing and comparing reinforcement learning algorithms, followed by OpenAI Baselines, which provided implementations of algorithms including DQN, ACKTR, and A2C. In October 2019, OpenAI demonstrated the practical application of these tools by training a robot hand to solve a Rubik's Cube. The company concluded the initiative by introducing Procgen, a benchmark for evaluating agent generalization across procedurally generated environments.

Wednesday, April 27, 2016

•
OpenAI Gym Beta OpenAI Blog •

Wednesday, May 24, 2017

•
OpenAI Baselines: DQN OpenAI Blog •

Friday, August 18, 2017

•
OpenAI Baselines: ACKTR & A2C OpenAI Blog •

Tuesday, October 15, 2019

•

Tuesday, December 03, 2019

•
Procgen Benchmark OpenAI Blog •