Stochastic Neural Networks for Hierarchical Reinforcement Learning | Synced

Authors proposed a general framework for learning a span of skills in a pre-training environment, which can be used in downstream tasks by training a high-level policy on top of the skills.