Return to Tradition: Learning Reliable Heuristics with Classical Machine Learning

Chen, D., Trevizan, F. and Thiébaux, S. To appear in Proc. of 34th Int. Conf. on Automated Planning and Scheduling (ICAPS).

We are working on the camera-ready of this paper and it will be available soon. Bellow is the abstract of this paper.

There has been a renewed interest in applying machine learning to planning due to recent developments in deep neural networks, with a lot of focus being placed on learning domain-dependent heuristics. However, current approaches for learning heuristics have yet to achieve competitive performance against domain-independent heuristics in several domains, and have poor overall performance. In this work, we construct novel graph representations of lifted planning tasks and use the WL algorithm to generate features from them. These features are used with classical machine learning methods such as Support Vector Machines and Gaussian Processes, which are both fast to train and evaluate. Our novel approach, WL-GOOSE, reliably learns heuristics from scratch and outperforms the h-FF heuristic. It also outperforms or ties with LAMA on 4 out of 10 domains. To our knowledge, the WL-GOOSE learned heuristics are the first to achieve these feats. Furthermore, we study the connections between our novel feature generation methods, previous theoretically flavoured learning architectures, and Description Logic features.

  • News and Highlights
    • GOOSE: First domain-independent method for learning heuristics based on lifted representations (AAAI'24).
    • CG-iLAO*: New planner for SSPs based on constraint generation (AAAI'24).
    • I was one of the organizers of GenPlan23 at NeurIPS.
    • First heuristic search algorithms for Multi-Objective Stochastic Planning (AAAI'23).
    • New admissible heuristics for Multi-Objective Deterministic Planning (ICAPS'22).
    • I gave an Early Career Researcher Spotlight talk at the IJCAI-20.
    • STRIPS-HGN: First algorithm capable of learning domain-independent heuristics from scratch (ICAPS'20).
    • ASNets: Learning generalized policies using neural nets (JAIR)
    • h-pom, h-roc, and i2-dual won the best paper award at ICAPS'17!
    • i-dual won the best paper award at ICAPS'16!
  • Word cloud of my papers
  • wordcloud
    large wordcloud