Automatic Discovery and Transfer of Task Hierarchies in Reinforcement Learning

  • Neville Mehta Oregon State University
  • Soumya Ray Case Western Reserve University
  • Prasad Tadepalli Oregon State University
  • Thomas Dietterich Oregon State University


Sequential decision tasks present many opportunities for the study of transfer learning. A principal one among them is the existence of multiple domains that share the same underlying causal structure for actions. We describe an approach that exploits this shared causal structure to discover a hierarchical task structure in a source domain, which in turn speeds up learning of task execution knowledge in a new target domain. Our approach is theoretically justified and compares favorably to manually designed task hierarchies in learning efficiency in the target domain. We demonstrate that causally motivated task hierarchies transfer more robustly than other kinds of detailed knowledge that depend on the idiosyncrasies of the source domain and are hence less transferable.

Author Biographies

Neville Mehta, Oregon State University
PhD Candidate, Department of Computer Science
Soumya Ray, Case Western Reserve University
Assistant Professor, Department of Electrical Engineering and Computer Science
Prasad Tadepalli, Oregon State University
Professor, Computer Science Department
Thomas Dietterich, Oregon State University
Professor and Director of Intelligent Systems Research
How to Cite
Mehta, N., Ray, S., Tadepalli, P., & Dietterich, T. (2011). Automatic Discovery and Transfer of Task Hierarchies in Reinforcement Learning. AI Magazine, 32(1), 35.