Multi-resolution Exploration in Continuous Spaces
نویسندگان
چکیده
The essence of exploration is acting to try to decrease uncertainty. We propose a new methodology for representing uncertainty in continuous-state control problems. Our approach, multi-resolution exploration (MRE), uses a hierarchical mapping to identify regions of the state space that would benefit from additional samples. We demonstrate MRE’s broad utility by using it to speed up learning in a prototypical model-based and value-based reinforcement-learning method. Empirical results show that MRE improves upon state-of-the-art exploration approaches.
منابع مشابه
Construction of continuous $g$-frames and continuous fusion frames
A generalization of the known results in fusion frames and $g$-frames theory to continuous fusion frames which defined by M. H. Faroughi and R. Ahmadi, is presented in this study. Continuous resolution of the identity (CRI) is introduced, a new family of CRI is constructed, and a number of reconstruction formulas are obtained. Also, new results are given on the duality of continuous fusion fram...
متن کاملMulitagent Reinforcement Learning in Stochastic Games with Continuous Action Spaces
We investigate the learning problem in stochastic games with continuous action spaces. We focus on repeated normal form games, and discuss issues in modelling mixed strategies and adapting learning algorithms in finite-action games to the continuous-action domain. We applied variable resolution techniques to two simple multi-agent reinforcement learning algorithms PHC and MinimaxQ. Preliminary ...
متن کاملEfficient Model-based Exploration in Continuous State-space Environments
OF THE DISSERTATION Efficient Model-based Exploration in Continuous State-space Environments by Ali Nouri Dissertation Director: Michael L. Littman The impetus for exploration in reinforcement learning (RL) is decreasing uncertainty about the environment for the purpose of better decision making. As such, exploration plays a crucial role in the efficiency of RL algorithms. In this dissertation,...
متن کاملPure exploration in finitely-armed and continuous-armed bandits
We consider the framework of stochastic multi-armed bandit problems and study the possibilities and limitations of forecasters that perform an on-line exploration of the arms. These forecasters are assessed in terms of their simple regret, a regret notion that captures the fact that exploration is only constrained by the number of available rounds (not necessarily known in advance), in contrast...
متن کاملPure Exploration for Multi-Armed Bandit Problems
We consider the framework of stochastic multi-armed bandit problems and study the possibilities and limitations of forecasters that perform an on-line exploration of the arms. These forecasters are assessed in terms of their simple regret, a regret notion that captures the fact that exploration is only constrained by the number of available rounds (not necessarily known in advance), in contrast...
متن کامل