Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsBlogPro
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges
  • Chrome Extension
  • Skill Manager

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Ogp Separating Quantum Inspired Optimization

ASecurity

Use when OGP separates classical heuristics from QAOA depth barriers.

3 stars
0 votes
0 copies
0 views
Added 10/3/2026
toolsgo

Security Analysis

A100/100

Scanned 10/3/2026

$npx -y skills add hiyenwong/ai_collection --skill ogp-separating-quantum-inspired-optimization --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Ogp Separating Quantum Inspired Optimization?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Ogp Separating Quantum Inspired Optimization
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/hiyenwong-ogp-separating-quantum-inspired-optimization/badge)](https://www.skillsdirectory.com/skills/hiyenwong-ogp-separating-quantum-inspired-optimization)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
Files
SKILL.md
---
name: ogp-separating-quantum-inspired-optimization
description: Use when OGP separates classical heuristics from QAOA depth barriers.
trigger: overlap gap property, OGP, QAOA vs mean-field MF-AOA, quantum-inspired classical algorithm limits, AMP obstruction, adiabatic to non-adiabatic QAOA parameter transition, clustered solution space, combinatorial optimization barrier
category: ai_collection
---

# OGP 分离量子与量子启发式优化算法

**来源**: arXiv:2609.35131 — Goh & Müller (DLR / Univ. Cologne), 2026-09-28

## 核心论断

问题具有 **Overlap Gap Property (OGP)** 时,解空间中 ε-最优解分裂为 Hamming 距离上不连通的簇(禁止重叠区间)。由此可以**刚性分离**两类算法:

1. **MF-AOA(量子启发经典算法)被阻塞**: 将 MF-AOA 的平均场自旋动力学嵌入**有限记忆 AMP** 框架(坐标式 Lipschitz 函数 f_t/F_t + 截断),直接继承 Gamarnik–Jagannath OGP 阻碍定理 → 无法任意接近最优。
2. **QAOA 可以穿越屏障**: Max-4-XORSAT 数值模拟显示 QAOA 超越 OGP 能量阈值,有限尺寸标度支持**超多项式**(亚指数)而非多项式深度依赖。

## 方法模式(可复用)

### 模式 1: 有限记忆 AMP 归约(证明经典启发式被 OGP 阻塞)

把一个经典启发式的迭代写成 AMP 递推 U^t = F_t(J(·, f_t(U^0..U^{t-1})), U^0..U^{t-1}),只需验证:
- f_t(0)=0、f_t/F_t 坐标式 Lipschitz(算子范数界:旋转矩阵正交 → ‖R_t(x)-R_t(y)‖ ≤ 2s(t)|x-y|,telescoping 得 K=2√T)
- 催化场项仿射 Lipschitz(max s²(1-s) = 4/27)

→ 任何能写成这种张量缩合 + 有限记忆 + 坐标截断形式的算法,都被 OGP 定理一网打尽。

### 模式 2: OGP 实例制备(branch-and-bound 扫描 ε)

对候选实例跑 branch-and-bound 收集 ε-优解集,计算重叠谱 q(σ,τ);逐次降 ε 直到谱出现**单个间隙** → OGP 实例;无任何 ε 产生间隙 → 无 OGP 对照实例。要点:对照实验用同一生成器产出,只差几何性质。

### 模式 3: 深度临界点检测(adiabatic → non-adiabatic 转变)

- 对单实例逐深度优化 (γ,β):浅深度最优参数呈线性绝热 schedule;到临界深度 p* 出现**不可微点**,全局最优跳到非绝热分支。
- 对齐技巧:把所有实例按 (p−p*, F_p−E_OGP) 平移后叠加 → 转变发生在 OGP 阈值附近略下方,是**实例共性**而非个体涨落。
- 优化器鲁棒性检查:COBYLA/BFGS/Nelder-Mead 换优化器、多初始化、Fourier 参数化都要复现同一 kink。

### 模式 4: 优化目标改变景观结构

优化 approximation ratio F_p 出现 kink;换成优化 ground-state overlap P_p 则 kink 消失。**可复用教训**: 变分参数的最优 schedule 定性依赖于所选优化目标——选择目标函数本身是算法设计自由度。TTS proxy = p/P_p 在有限深度取最小(28–52 层),盲目加深不如在 TTS 最优点停。

### 模式 5: 资源估计桥接算法-硬件鸿沟

两条曲线判定硬件可达性:
- 屏障深度 p_OGP(N)(数值外推,N=50 时 p≈250,线性外推为乐观下界)
- 相干极限深度 p_T2 = T2 / (D_p·t_2Q),D_p = G_p/(Δ+1)(并行度折算)

超导参数下 p_T2 = 13–33 ≪ p_OGP → **当前硬件在到达 OGP 屏障之前就耗尽相干时间**。保真度模型 F_2Q = (1−e_2Q)^G 随深度指数衰减。任何"量子优势宣称"都应同时报告 p_barrier 与 p_T2 的差距。

## 实验结果速查

| 观察 | 数值 |
|------|------|
| N=15 代表实例临界深度 | p* ≈ 28 |
| TTS 最优深度 (N=15→21) | p = 28 → 52 |
| N=50 外推屏障深度 | p ≈ 250(线性外推,乐观)|
| 超多项式拟合优势 (N=27-29) | RMSE_poly/RMSE_spoly ≈ 10.95 |
| 小尺寸 (N=23-26) | RMSE 比率 0.44(多项式略优,无法区分)|
| IBM Heron 相干极限 | p_T2 = 13–33 ≪ 250 |
| MF-AOA Lipschitz 常数 | K ≤ 2√T(与 N 无关)|

## 应用边界

- 稀疏图上对数深度 QAOA 已知被 OGP 限制(Chou et al.);本文给出的是**穿越屏障所需深度**的第一个数值刻画。
- MF-AOA 阻碍结论 conditional on H_N 上的 OGP 扩展假设(Conjecture 3.2 of Gamarnik-Jagannath)。
- N=50 是经典模拟开始失效的交叉点,也是首个可能的量子优势实验规模。

## 相关技能

- `qaoa-landscape-audit` — QAOA 变分景观审计
- `comet-constraint-preserving-qaoa` — 约束保持 QAOA
- `qaoa-interaction-threshold` — QAOA 模拟复杂度阈值
- `quantum-mcts-fixed-confidence` — 量子 vs 经典混合成本切换

## 参考文献

1. Goh & Müller, arXiv:2609.35131 (2026)
2. Gamarnik & Jagannath, Ann. Probab. 49(1) 180 (2021) — AMP 的 OGP 阻碍定理
3. Chou, Love, Sandhu, Shi, ICALP 2022 — 局域量子算法极限
4. Boulebnane & Montanaro, PRX Quantum 5, 030348 (2024) — k-SAT QAOA 指数标度
5. Misra-Spieldenner et al., PRX Quantum 4, 030335 (2023) — MF-AOA 原始定义
6. Zhou, Wang, Choi, Pichler, Lukin, PRX 10, 021067 (2020) — Fourier 参数化

Attribution

hiyenwonghiyenwong
View sourceSee grades on GitHubMore from hiyenwong →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

ucoz-landing-skill

Create and edit uCoz homepage landing pages via MCP: custom templates, hero sections, lead forms, navigation menus, SEO, and responsive layout. Includes a visual design system (style selection, layout/grid, section recipes, typography/spacing, color tokens, component states, icons, modern CSS/JS, motion, imagery, social proof, copy/voice, accessibility). Uses ucoz-mcp tools for templates, site file uploads, and site modules.

107 votes

Paperclip

Interact with the Paperclip control plane API for task coordination and governance. Use when checking assignments, updating issue status, posting comments, delegating work, managing routines, or calling Paperclip API endpoints.

953191 votes

Pptx

Presentation toolkit (.pptx). Create/edit slides, layouts, content, speaker notes, comments, for programmatic presentation creation and modification.

471861 votes

Daw Music

Digital Audio Workstation usage, music composition, interactive music systems, and game audio implementation for immersive soundscapes.

761 votes

Instantly Rdsthomas Mission Control

Instantly.ai cold email outreach API - manage campaigns, leads, accounts, and analytics. Use for cold email automation, lead management, campaign creation/monitoring, and email account warmup.

761 votes
View all in tools →