OpenAI Baselines: ACKTR & A2C
OpenAI Blog
OpenAI released two new reinforcement learning algorithm implementations: A2C, a synchronous version of A3C that performs equally well, and ACKTR, which improves sample efficiency over both TRPO and A2C. ACKTR requires only marginally more computation per update than A2C while achieving better sample efficiency. These implementations are now available in OpenAI Baselines, giving researchers easier access to these algorithms for training reinforcement learning agents.
Why it matters
We’re releasing two new OpenAI Baselines implementations: ACKTR and A2C. A2C is a synchronous, deterministic variant of Asynchronous Advantage Actor Critic (A3C) which we’ve found gives equal performance. ACKTR is a more sample-efficient reinforcement learning algorithm than TRPO and A2C, and requires only slightly more computation than A2C per update.
