量子干涉作为组合优化的提议机制
Quantum Interference as a Proposal Mechanism for Combinatorial Optimization
AI总结：
该研究提出量子干涉提议搜索（QIPS）方法，采用种子条件量子电路生成局域干涉图案作为优化提议分布，在6个基准上验证其作为资源高效组合优化提议机制的竞争力。
AI中文摘要：
量子干涉提议搜索（QIPS）采用种子条件量子电路生成局域干涉图案，作为QUBO/Ising优化的有限次采样提议分布。从这些分布中采样候选n_b位字符串，经经典评分后用于更新低能解的精英前沿。QIPS使用固定的两层门控电路架构，每个电路采样100次，希尔伯特空间维度随2^n_b增长。在6个基准系列（n_b为18至29）中，QIPS与匹配的经典对照（保留相同搜索循环、前沿更新规则和提议预算，总提议数与n_b成正比）相比保持了有竞争力的进展，性能通过top-K覆盖率、命中率、多重性、希尔伯特空间覆盖率和二元秩指标评估。结果表明，局域量子干涉是计算量子优化的资源高效提议机制。
英文摘要：
Quantum Interference Proposal Search (QIPS) uses seed-conditioned quantum circuits to generate localized interference patterns as finite-shot proposal distributions for QUBO/Ising optimization. Candidate $n_b$-bit strings are sampled from these distributions, scored classically and used to update an elite frontier of low-energy solutions. QIPS uses a fixed two-layer gate-based circuit architecture with 100 shots per circuit while the Hilbert-space dimension grows as $2^{n_b}$. Across six benchmark families with $18 \le n_b \le 29$, QIPS maintains competitive progress relative to a matched classical control that preserves the same search loop, frontier update rule and proposal budget, with total proposals proportional to $n_b$. Performance is assessed using top-$K$ coverage, hit rate, multiplicity, Hilbert-space coverage and dyadic-rank metrics. The results identify localized quantum interference as a resource-efficient proposal mechanism for computational quantum optimization.
