Uncertainty-sensitive Learning and Planning with Ensembles
arXiv:1912.09996
Abstract
We propose a reinforcement learning framework for discrete environments in which an agent makes both strategic and tactical decisions. The former manifests itself through the use of value function, while the latter is powered by a tree search planner. These tools complement each other. The planning module performs a local \textit{what-if} analysis, which allows to avoid tactical pitfalls and boost backups of the value function. The value function, being global in nature, compensates for inherent locality of the planner. In order to further solidify this synergy, we introduce an exploration mechanism with two distinctive components: uncertainty modelling and risk measurement. To model the uncertainty we use value function ensembles, and to reflect risk we use propose several functionals that summarize the implied by the ensemble. We show that our method performs well on hard exploration environments: Deep-sea, toy Montezuma's Revenge, and Sokoban. In all the cases, we obtain speed-up in learning and boost in performance.
References in corpus (15)
- Deep Exploration via Bootstrapped DQN
- Model-Based Reinforcement Learning for Atari
- Hindsight Experience Replay
- Continuous Deep Q-Learning with Model-based Acceleration
- Model-Ensemble Trust-Region Policy Optimization
- Go-Explore: a New Approach for Hard-Exploration Problems
- Quantifying Generalization in Reinforcement Learning
- Reinforcement and Imitation Learning via Interactive No-Regret Learning
- Randomized Prior Functions for Deep Reinforcement Learning
- Learning model-based planning from scratch
- Plan Online, Learn Offline: Efficient Learning and Exploration via Model-Based Control
- The Uncertainty Bellman Equation and Exploration
- Successor Uncertainties: Exploration and Uncertainty in Temporal Difference Learning
- Improving Hearthstone AI by Combining MCTS and Supervised Learning Algorithms
- Adaptive Online Planning for Continual Lifelong Learning