EVENT

Talk on "Machine-Learning-Based Meta Algorithm Design for Solvers" by John Lu (University of Waterloo)

We are pleased to inform you about the upcoming seminar by John Lu (University of Waterloo) titled:"Machine-Learning-Based Meta Algorithm Design for Solvers" Everyone interested is cordially invited to attend!

Title:

Machine-Learning-Based Meta Algorithm Design for Solvers

Abstract:

For NP-hard problems such as SAT, SMT, and model checking, no known single algorithm efficiently solves all instances. In practice, different solvers often exhibit unique strengths and weaknesses over various instance families. To leverage these complementary strengths, machine learning has been successfully applied to construct meta-solvers that combine off-the-shelf solvers, for the purpose of optimizing performance for specific use cases. This talk presents two studies in this area: one investigates graph-based circuit features for algorithm selection in hardware model checking, while the other applies Monte Carlo Tree Search to synthesize SMT solving strategies. The talk concludes with a discussion of this topic from a learning-theoretical perspective.

Speaker:

John Lu (University of Waterloo)

Time/Date:

16:30- July 8 (Tuesday), 2025

Place:

Room 1310A, NII & online (zoom)

Contact:

If you would like to join, please contact by email.
Email :kphalakarn[at]nii.ac.jp