Information asymmetry in KL-regularized RL
arXiv:1905.01240
Abstract
Many real world tasks exhibit rich structure that is repeated across different parts of the state space or in time. In this work we study the possibility of leveraging such repeated structure to speed up and regularize learning. We start from the KL regularized expected reward objective which introduces an additional component, a default policy. Instead of relying on a fixed default policy, we learn it from data. But crucially, we restrict the amount of information the default policy receives, forcing it to learn reusable behaviors that help the policy learn faster. We formalize this strategy and discuss connections to information bottleneck approaches and to the variational EM algorithm. We present empirical results in both discrete and continuous action domains and demonstrate that, for certain tasks, learning a default policy alongside the policy can significantly speed up and improve learning.
Accepted as a conference paper at ICLR 2019
Cited by in corpus (13)
- Learning Agile Soccer Skills for a Bipedal Robot with Deep Reinforcement Learning
- Keep Doing What Worked: Behavioral Modelling Priors for Offline Reinforcement Learning
- Exploiting Hierarchy for Learning and Transfer in KL-regularized RL
- Behavior Priors for Efficient Reinforcement Learning
- Learning Dexterous Manipulation from Suboptimal Experts
- Entropy Regularization with Discounted Future State Distribution in Policy Gradient Methods
- Marginalized State Distribution Entropy Regularization in Policy Optimization
- Imagined Value Gradients: Model-Based Policy Optimization with Transferable Latent Dynamics Models
- An Empowerment-based Solution to Robotic Manipulation Tasks with Sparse Rewards
- Ground States of Quantum Many Body Lattice Models via Reinforcement Learning
- Evaluating model-based planning and planner amortization for continuous control
- Towards an Understanding of Default Policies in Multitask Policy Optimization
- Conditioning of Reinforcement Learning Agents and its Policy Regularization Application