Distributed training using policy off actor explicit reinforcement learning

摘要

A method and system comprising a computer program encoded on a computer storage medium for training an action selective neural network that is used to select an action performed by interacting with an environment And apparatus.In one embodiment, the system includes a plurality of actor computing units and a plurality of Larner computing units.The actor computing unit uses the enhanced learning technique to generate the trajectory of the experience tuple used by the Larner computing unit to update the parameters of the Larner action selective neural network.Reinforcement learning technology may be a policy offensive reinforcement learning technique.