Learning for Planning and Problem Solving
Most learning research concerns classification. Research in learning and planning and problem solving focuses on improving the performance of an AI planning or problem solving system through experience. Our work has focussed on integrating explanation-based learning (EBL) and inductive learning (specifically ILP) to improve the efficiency (speedup learning) and solution-quality for planning and problem solving systems by solving sample problems and learning heuristics that avoid backtracking or sub-optimal solutions.

Our work has focused on two systems:

  • SCOPE: Learning control rules for partial-order planning to improve efficiency and plan quality
  • DOLPHIN: Learning clause-selection rules for dynamic optimization of logic programs
Learning Behavior Characterizations for Novelty Search Elliot Meyerson, Joel Lehman and Risto Miikkulainen To Appear In Proceedings of the Genetic and Evolutionary Computation Conference (GECCO 2016),... 2016

Object-Model Transfer in the General Video Game Domain Alexander Braylan, Risto Miikkulainen To Appear In Proceedings of the Twelfth AAAI Conference on Artificial Intelligence and Interactiv... 2016

Learning Polarity from Structure in SAT Bryan Silverthorn and Risto Miikkulainen In Theory and Applications of Satisfiability Testing (SAT), 2011. (extended abstract). 2011

Latent Class Models for Algorithm Portfolio Methods Bryan Silverthorn and Risto Miikkulainen In Proceedings of the Twenty-Fourth AAAI Conference on Artificial Intelligence, 2010. 2010

Alexander Braylan braylan [at] cs utexas edu
Bryan Silverthorn Ph.D. Alumni bsilvert [at] cs utexas edu

The borg project includes a practical algorithm...