Knowledge Extraction from Reinforcement Learning
This chapter is concerned with knowledge extraction from reinforcement learners. It addresses two approaches towards knowledge extraction: the extraction of explicit, symbolic rules from neural reinforcement learners, and the extraction of complete plans from such learners. The advantages of such knowledge extraction include (1) the improvement of learning (especially with the rule extraction approach), and (2) the improvement of the usability of results of learning.
KeywordsReinforcement Learner Bottom Level Sequential Decision Rule Extraction Knowledge Extraction
Unable to display preview. Download preview PDF.
- 1.Fu, L.M. (1991), “Rule learning by searching on adapted nets,” Proc. of AAAI’91, pp. 590–595.Google Scholar
- 3.Lin, L. (1992), “Self-improving reactive agents based on reinforcement learning, planning, and teaching,” Machine Learning, vol. 8, pp. 293–321.Google Scholar
- 4.Maclin, R. and Shavlik, J. (1994), “Incorporating advice into agents that learn from reinforcements,” Proc. of AAAI-9.4, Morgan Kaufmann, San Meteo, CA.Google Scholar
- 7.Sun, R. (1997), “Multi-agent approaches toward reinforcement learning,” TRCS-97–0028, University of AlabamaGoogle Scholar
- 8.Sun, R. and Peterson, T. (1997), “A hybrid model for learning sequential navigation,” Proc. of IEEE International Symposium on Computational Intelligence in Robotics and Automation (CIRA’97), IEEE Press, Monterey, CA, pp. 234–239.Google Scholar
- 10.Sun, R., Peterson, T., and Merrill, E. (1999), “A hybrid architecture for situated learning of reactive sequential decision making,” Applied Intelligence,in press.Google Scholar
- 11.Sun, R. and Sessions, C. (1998), “Extracting plans from reinforcement learners,” in L. Xu, L. Chan, I. King, and A. Fu (Eds.), Proceedings of the 1998 International Symposium on Intelligent Data Engineering and Learning (IDEAL’98), Springer-Verlag, pp. 243–248.Google Scholar
- 12.Sun, R. and Sessions, C. (1998), “Learning to plan probabilistically from neural networks,” Proceedings of IEEE International Conference on Neural Networks, Anchorage, Alaska. May 4–9, IEEE Press, Piscataway, NJ, pp. 1–6.Google Scholar
- 13.Sutton, R. (1990), “Integrated architectures for learning, planning, and reacting based on approximating dynamic programming,” Proc. of Seventh International Conference on Machine Learning, Morgan Kaufmann, San Mateo, CA.Google Scholar
- 15.Towell, G. and Shavlik, J. (1993), “Extracting refined rules from Knowledge-Based Neural Networks,” Machine Learning, vol. 13, no. 1, pp. 71–101.Google Scholar
- 16.Watkins, C. (1989), Learning with Delayed Rewards, Ph.D. Thesis, Cambridge University, Cambridge, U.K.Google Scholar