ERA is in the process of being migrated to Scholaris, a Canadian shared institutional repository service (https://scholaris.ca). Deposits and changes to existing ERA items and collections are frozen until migration is complete. Please contact erahelp@ualberta.ca for further assistance
- 681 views
- 758 downloads
Reinforcement Learning Algorithms for MDPs
-
- Author(s) / Creator(s)
-
Technical report TR09-13. This article presents a survey of reinforcement learning algorithms for Markov Decision Processes (MDP). In the first half of the article, the problem of value estimation is considered. Here we start by describing the idea of bootstrapping and temporal difference learning. Next, we compare incremental and batch algorithmic variants and discuss the impact of the choice of the function approximation method on the success of learning. In the second half, we describe methods that target the problem of learning to control an MDP. Here online and active learning are discussed first, followed by a description of direct and actor-critic methods. | TRID-ID TR09-13
-
- Date created
- 2009
-
- Subjects / Keywords
-
- Artificial Intelligence
- Monte-Carlo methods
- Reinforcement learning
- Actor-critic methods
- Stochastic approximation
- Markov decision processes
- Active learning
- Overfitting
- L:east-sqares methods
- Temporal difference learning
- Simulations
- Policy gradient
- Two-timescale stochastic approximation
- Q-learning
- Online learning
- Function approximation
- Natural gradient
- PAC-learning
- Machine Learning
- Planning
- Stochastic gradient methods
- Stimulation optimization
- Bias-variance tradeoff
-
- Type of Item
- Report
-
- License
- Attribution 3.0 International