论文信息 - Towards Informed Reinforcement Learning

Towards Informed Reinforcement Learning

In this paper, we introduce the ideas of Informed Reinforcement Learning, an extension of Relational Reinforcement Learning, in which partial world models are being learned, in order to improve convergence to the optimal policy, using goal-oriented reasoning. Furthermore, an algorithm is presented to enhance an RRL-agent with reasoning capabilities.

Maurice Bruynooghe | Jan Ramon | Tom Croonenborghs

[1] Saso Dzeroski,et al. Integrating Experimentation and Guidance in Relational Reinforcement Learning , 2002, ICML.

[2] Kurt Driessens,et al. Relational Instance Based Regression for Relational Reinforcement Learning , 2003, ICML.

[3] Xin Wang,et al. Model-based Policy Gradient Reinforcement Learning , 2003, ICML.

[4] Luc De Raedt,et al. Towards Discovering Structural Signatures of Protein Folds Based on Logical Hidden Markov Models , 2003, Pacific Symposium on Biocomputing.

[5] Daniel S. Weld. Solving Relational MDPs with First-Order Machine Learning , 2004 .

[6] Luc De Raedt,et al. Bellman goes relational , 2004, ICML.

[7] De,et al. Relational Reinforcement Learning , 2001, Encyclopedia of Machine Learning and Data Mining.

[8] Luc De Raedt,et al. Towards Combining Inductive Logic Programming with Bayesian Networks , 2001, ILP.

[9] Craig Boutilier,et al. Symbolic Dynamic Programming for First-Order MDPs , 2001, IJCAI.

[10] Jonathan Baxter. KnightCap : A chess program that learns by combining TD ( ) with game-tree search , 1998 .

[11] Andrew Tridgell,et al. KnightCap: A Chess Programm That Learns by Combining TD(lambda) with Game-Tree Search , 1998, ICML.

[12] Natalia,et al. Applying probabilistic rules to relational worlds , 2003 .