“Digital Annealer” Technology for Efficiently Solving Combinatorial Optimization Problems and Its Applications in Manufacturing and Materials Science

DOI

Bibliographic Information

Other Title
  • 組合せ最適化問題を高速に解く「デジタルアニーラ」技術と製造・材料科学への応用

Abstract

Digital Annealer is a dedicated architecture for quickly solving combinatorial optimization problems mapped to an Ising model. Its characteristic is being able to handle fully connected problems at high-resolution coupling strength. Although Digital Annealer's basic search mechanism is based on the Markov-Chain Monte-Carlo method, several techniques such as parallel trials, escaping from local minimum, and replica exchange are implemented inside the hardware for further speedup. This paper describes the operating principle of the Digital Annealer and benchmarks the performance of the maximum cut, minimum cut and quadratic assignment problem for the second-generation chip with 8,192 bits. As examples, formulation and evaluation results of job shop scheduling and crystal structural similarity are presented to show that the fields of manufacturing and material science are promising application areas for the Digital Annealer.

Keywords

Details 詳細情報について

Report a problem

Back to top