/hierarchical-deep-RL

Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstractions and Intrinsic Motivation

Primary LanguageLua

Code for h-DQN, NIPS 2016

  • Use the synthetic branch for the stochastic decision process example
  • Use the metanet branch for Atari. Pre-train the network using the iclr16_basicsubgoal branch before doing this and load this network using the metanet branch to train the full model.