Scholarlink

<p>Combinatorial optimization underpins applications in artificial intelligence, logistics, and network design, yet classical techniques such as greedy search and dynamic programming struggle to balance efficiency and solution quality at scale. We present a probabilistic framework that embeds true random number generators based on spin-transfer-torque magnetic tunnel junctions into a greedy solver. Intrinsic stochastic switching enables configurable random number distributions, which we use to inject controlled randomness via a temperature parameter that interpolates between deterministic and stochastic choices, balancing exploration and exploitation. Applied to the traveling salesman problem, the framework yields high-quality tours and outperforms simulated annealing and genetic algorithms in solution quality and convergence speed. In larger instances with up to 70 cities, it maintains its advantage, reaching near-optimal solutions with fewer iterations and reduced computational cost. These results show that hardware true randomness with tunable statistics can improve heuristic search and motivate integrated, energy-efficient probabilistic hardware for scalable optimization.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Probabilistic greedy algorithm solver using magnetic tunneling junctions for traveling salesman problem

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月4日

查看全文
  • Ran Zhang,
  • Xiaohan Li,
  • Caihua Wan,
  • Raik Hoffmann,
  • Meike Hindenberg,
  • Yingqian Xu,
  • Shiqiang Liu,
  • Dehao Kong,
  • Shilong Xiong,
  • Shikun He,
  • Alptekin Vardar,
  • Qiang Dai,
  • Junlu Gong,
  • Yihui Sun,
  • Zejie Zheng,
  • Thomas Kämpfe,
  • Guoqiang Yu,
  • Xiufeng Han

摘要

Combinatorial optimization underpins applications in artificial intelligence, logistics, and network design, yet classical techniques such as greedy search and dynamic programming struggle to balance efficiency and solution quality at scale. We present a probabilistic framework that embeds true random number generators based on spin-transfer-torque magnetic tunnel junctions into a greedy solver. Intrinsic stochastic switching enables configurable random number distributions, which we use to inject controlled randomness via a temperature parameter that interpolates between deterministic and stochastic choices, balancing exploration and exploitation. Applied to the traveling salesman problem, the framework yields high-quality tours and outperforms simulated annealing and genetic algorithms in solution quality and convergence speed. In larger instances with up to 70 cities, it maintains its advantage, reaching near-optimal solutions with fewer iterations and reduced computational cost. These results show that hardware true randomness with tunable statistics can improve heuristic search and motivate integrated, energy-efficient probabilistic hardware for scalable optimization.