Presentation Information
[9p-N106-5]A Comparative Study of Simulated Quantum Annealing with Extraction-Type Majority Voting Logic and OpenJij's SQASampler for Combinatorial Optimization Problems
〇Yu Zhao1, Koki Awaya1, Moe Shimada1, Jun-ichi Shirakashi1 (1.Tokyo Univ. Agr. and Tech.)
Keywords:
Ising machine,Simulated Quantum Annealing,Combinatorial optimization problems
Combinatorial optimization problems have attracted significant attention in the context of quantum annealing and its quantum-inspired counterparts. Among these, quantum-inspired Ising machines have shown promise in overcoming the physical limitations of quantum hardware while efficiently solving large-scale optimization tasks. Simulated quantum annealing (SQA), a quantum-inspired algorithm based on path-integral Monte Carlo methods, is frequently applied to spin glass problems such as the Sherrington-Kirkpatrick (SK) model. OpenJij is an open-source framework that provides implementations of SQA and simulated annealing (SA) through dedicated samplers (SQASampler and SASampler) with support for flexible scheduling on CPU backends. In this study, we use our own implementation of SQA, which allows flexible adjustment of various parameters. We also extend this approach with a hybrid variant that incorporates extraction-type majority voting logic (E-MVL) to enhance performance. We evaluate it against OpenJij to perform SQA under equivalent experimental conditions, including identical annealing schedules and coupling constants for the Sherrington-Kirkpatrick (SK) model.