A3C
Asynchronous Advantage Actor-Critic (A3C) is a distributed reinforcement learning algorithm proposed by DeepMind. It allows multiple workers to each maintain a local copy of the network and asynchronously push gradients to a shared global network. A3C is an early representative of the Actor-Learner decoupled architecture.
Timeline (last 90 days)
Actor-Learner decoupling is a common approach in distributed reinforcement learning, and architectures such as A3C, IMPALA, and Ape-X are all based on this concept.
A3C allows multiple workers to each maintain a copy of the network and asynchronously push gradients to the global network.
All Facts (2)
Actor-Learner decoupling is a common approach in distributed reinforcement learning, and architectures such as A3C, IMPALA, and Ape-X are all based on this concept.
50%UnverifiedA3C allows multiple workers to each maintain a copy of the network and asynchronously push gradients to the global network.
50%