AI 中文总结
该研究提出时间维度交换耦合（TEC）方案，无需复制副本硬件即可提升伊辛机采样效率，在顶点数达2000的MaxCut问题中显著加快收敛速度，经SPICE仿真验证硬件可行性。
AI 中文摘要
传统概率伊辛机常存在构型空间探索效率低下的问题，而基于副本的量子蒙特卡洛方法虽能缓解采样瓶颈，但需付出巨大硬件开销。本文提出一种时间维度交换耦合（TEC），用单个p位网络连续自旋构型间的时间交换相互作用替代空间副本耦合，无需复制副本硬件即可提升采样效率。低温下，反铁磁TEC扩大采样范围；高温下，铁磁TEC稳定最优状态。针对顶点数达2000的最大割（MaxCut）问题，TEC显著加快收敛速度，SPICE仿真验证了硬件可行性，为在现有伊辛机上增强组合优化提供了可扩展、硬件高效的策略。
英文摘要
Conventional probabilistic Ising machines often suffer from inefficient exploration of configuration space, while replica-based quantum Monte Carlo methods reduce sampling bottlenecks at the cost of large hardware overhead. Here we propose a time-dimensional exchange coupling (TEC) that replaces spatial replica coupling with a temporal exchange interaction between successive spin configurations of a single p-bit network. This TEC improves sampling efficiency without duplicating replica hardware. At low temperatures, antiferromagnetic TEC expands sampling range; at high temperatures, ferromagnetic TEC stabilizes the optimal state. For MaxCut problems with up to 2000 vertices, TEC significantly accelerates convergence speed. SPICE simulations confirm hardware feasibility. TEC offers a scalable, hardware-efficient strategy to enhance combinatorial optimization on existing Ising machines.
Comments8 pages, 4 figures
