Fitted Q-Learning for Relational Domains

Author(s)

Srijita Das, University of Texas at DallasFollow
Sriraam Natarajan, University of Texas at DallasFollow
Kaushik Roy, University of South Carolina - ColumbiaFollow
Ron Parr, Duke UniversityFollow
Kristian Kersting, TU DarmstadtFollow

ORCID iD

https://orcid.org/0000-0001-6610-7845

Document Type

Poster

Subject Area(s)

Reinforcement Learning, First Order Logic

Abstract

We consider the problem of Approximate Dynamic Program- ming in relational domains. Inspired by the success of fit- ted Q-learning methods in propositional settings, we develop the first relational fitted Q-learning algorithms by represent- ing the value function and Bellman residuals. When we fit the Q-functions, we show how the two steps of Bellman op- erator; application and projection steps can be performed us- ing a gradient-boosting technique. Our proposed framework performs reasonably well on standard domains without using domain models and using fewer training trajectories.

Digital Object Identifier (DOI)

arXiv:2006.05595

Publication Info

Knowledge representation and Reasoning 2020, Fall 2020.

© The Authors, 2020

APA Citation

Das, S., Natarajan, S., Roy, K., Parr, R., & Kersting, K. (2020). Fitted Q-Learning for Relational Domains. arXiv preprint arXiv:2006.05595.