[{"data":1,"prerenderedAt":497},["ShallowReactive",2],{"content-query-tYoW8Zbv8M":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"category":13,"body":14,"_type":491,"_id":492,"_source":493,"_file":494,"_stem":495,"_extension":496},"/technology-blogs/zh/2026-1-15","zh",false,"","开源之夏｜孙文杰：基于MindSpore Quantum的对称性规约与分岔分析算法实践","针对复杂物理系统（如Z2格点规范场）基态求解中的“结构盲点”痛点，提出了一种基于对称性的问题规约方法与同伦延拓分岔分析框架","2026-1-15","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2025/06/06/1a18a46ef03442ea8f8d83ba33b0a7af.png","technology-blogs","开发者说",{"type":15,"children":16,"toc":480},"root",[17,25,36,46,54,89,97,105,109,115,124,184,188,194,203,211,218,225,233,241,248,256,264,271,279,287,294,300,343,351,359,367,385,393,411,419,427],{"type":18,"tag":19,"props":20,"children":22},"element","h1",{"id":21},"开源之夏孙文杰基于mindspore-quantum的对称性规约与分岔分析算法实践",[23],{"type":24,"value":8},"text",{"type":18,"tag":26,"props":27,"children":29},"div",{"style":28},"text-align: center;",[30],{"type":18,"tag":31,"props":32,"children":35},"img",{"src":33,"style":34,"alt":7},"/category/information/technology-blogs/banner/2026-1-15/1.jpg","display: block;margin: 0 auto;max-width:70%",[],{"type":18,"tag":37,"props":38,"children":39},"p",{},[40],{"type":18,"tag":41,"props":42,"children":43},"strong",{},[44],{"type":24,"value":45},"# 01",{"type":18,"tag":37,"props":47,"children":48},{},[49],{"type":18,"tag":41,"props":50,"children":51},{},[52],{"type":24,"value":53},"项目介绍",{"type":18,"tag":55,"props":56,"children":57},"ul",{},[58,64,69],{"type":18,"tag":59,"props":60,"children":61},"li",{},[62],{"type":24,"value":63},"项目名称：结构感知的量子启发优化算法",{"type":18,"tag":59,"props":65,"children":66},{},[67],{"type":24,"value":68},"项目描述：本项目针对复杂物理系统（如Z2格点规范场）基态求解中的“结构盲点”痛点，提出了一种基于对称性的问题规约方法与同伦延拓分岔分析框架。通过利用问题固有的对称性约束将高阶HUBO问题降阶，并结合MindSpore Quantum实现了改进的LQA算法与高效分岔图可视化，显著提升了组合优化问题的求解效率与解空间探索能力。",{"type":18,"tag":59,"props":70,"children":71},{},[72,77,81],{"type":18,"tag":41,"props":73,"children":74},{},[75],{"type":24,"value":76},"项目源码链接：",{"type":18,"tag":78,"props":79,"children":80},"br",{},[],{"type":18,"tag":82,"props":83,"children":87},"a",{"href":84,"rel":85},"https://gitee.com/mindspore/mindquantum/tree/research/summer_ospp/2025/25c6d0296",[86],"nofollow",[88],{"type":24,"value":84},{"type":18,"tag":37,"props":90,"children":91},{},[92],{"type":18,"tag":41,"props":93,"children":94},{},[95],{"type":24,"value":96},"# 02",{"type":18,"tag":37,"props":98,"children":99},{},[100],{"type":18,"tag":41,"props":101,"children":102},{},[103],{"type":24,"value":104},"技术深潜：实现路径与核心创新",{"type":18,"tag":106,"props":107,"children":108},"h3",{"id":7},[],{"type":18,"tag":106,"props":110,"children":112},{"id":111},"_1技术栈与系统架构",[113],{"type":24,"value":114},"1、技术栈与系统架构",{"type":18,"tag":106,"props":116,"children":118},{"id":117},"项目以mindspore-quantum-作为核心开发框架结合-networkx-处理图论计算利用-python-实现算法流程与数据可视化整体架构分为四层",[119],{"type":18,"tag":41,"props":120,"children":121},{},[122],{"type":24,"value":123},"项目以MindSpore Quantum 作为核心开发框架，结合 NetworkX 处理图论计算，利用 Python 实现算法流程与数据可视化。整体架构分为四层：",{"type":18,"tag":55,"props":125,"children":126},{},[127,142,156,170],{"type":18,"tag":59,"props":128,"children":129},{},[130,135,137],{"type":18,"tag":41,"props":131,"children":132},{},[133],{"type":24,"value":134},"图模型构建层：",{"type":24,"value":136}," ",{"type":18,"tag":41,"props":138,"children":139},{},[140],{"type":24,"value":141},"将物理问题（HUBO/Ising）抽象为标准化的图数据结构。",{"type":18,"tag":59,"props":143,"children":144},{},[145,150,151],{"type":18,"tag":41,"props":146,"children":147},{},[148],{"type":24,"value":149},"核心算法层（对称性计****算）：",{"type":24,"value":136},{"type":18,"tag":41,"props":152,"children":153},{},[154],{"type":24,"value":155},"基于Kruskal算法生成最小生成森林，识别弦并构建环路，计算系统的局部规范对称性。",{"type":18,"tag":59,"props":157,"children":158},{},[159,164,165],{"type":18,"tag":41,"props":160,"children":161},{},[162],{"type":24,"value":163},"问题规约与求解层：",{"type":24,"value":136},{"type":18,"tag":41,"props":166,"children":167},{},[168],{"type":24,"value":169},"利用对称性构造对偶变换算子，将高阶含约束的原始哈密顿量映射为低阶无约束的对偶Ising模型，并使用改进后的LQA（局域量子退火）算法求解。",{"type":18,"tag":59,"props":171,"children":172},{},[173,178,179],{"type":18,"tag":41,"props":174,"children":175},{},[176],{"type":24,"value":177},"分岔分析与可视化层：",{"type":24,"value":136},{"type":18,"tag":41,"props":180,"children":181},{},[182],{"type":24,"value":183},"使用同伦延拓法追踪解路径，检测分岔点，并绘制PCA轨迹图与分岔图。",{"type":18,"tag":106,"props":185,"children":187},{"id":186},"_1",[],{"type":18,"tag":106,"props":189,"children":191},{"id":190},"_2关键实现步骤详解",[192],{"type":24,"value":193},"2、关键实现步骤详解",{"type":18,"tag":106,"props":195,"children":197},{"id":196},"步骤一对称性计算与模块化编程",[198],{"type":18,"tag":41,"props":199,"children":200},{},[201],{"type":24,"value":202},"步骤一：对称性计算与模块化编程",{"type":18,"tag":37,"props":204,"children":205},{},[206],{"type":18,"tag":41,"props":207,"children":208},{},[209],{"type":24,"value":210},"实现对称性搜索函数函数，利用循环基理论和Kruskal算法，在复杂度内识别出图的所有基础对称性结构。",{"type":18,"tag":26,"props":212,"children":213},{"style":28},[214],{"type":18,"tag":31,"props":215,"children":217},{"src":216,"style":34,"alt":7},"/category/information/technology-blogs/banner/2026-1-15/2.jpg",[],{"type":18,"tag":26,"props":219,"children":220},{"style":28},[221],{"type":18,"tag":31,"props":222,"children":224},{"src":223,"style":34,"alt":7},"/category/information/technology-blogs/banner/2026-1-15/3.jpg",[],{"type":18,"tag":37,"props":226,"children":227},{},[228],{"type":18,"tag":41,"props":229,"children":230},{},[231],{"type":24,"value":232},"步骤二： 对偶变换与LQA改进",{"type":18,"tag":37,"props":234,"children":235},{},[236],{"type":18,"tag":41,"props":237,"children":238},{},[239],{"type":24,"value":240},"基于Wegner对偶变换理论，将原始的四阶磁场项和一阶电场项转换为对偶空间中的二阶和一阶项，去除了复杂的局部高斯定律约束，从而实现算法降阶。",{"type":18,"tag":26,"props":242,"children":243},{"style":28},[244],{"type":18,"tag":31,"props":245,"children":247},{"src":246,"style":34,"alt":7},"/category/information/technology-blogs/banner/2026-1-15/4.jpg",[],{"type":18,"tag":37,"props":249,"children":250},{},[251],{"type":18,"tag":41,"props":252,"children":253},{},[254],{"type":24,"value":255},"步骤三： 同伦延拓分岔搜索",{"type":18,"tag":37,"props":257,"children":258},{},[259],{"type":18,"tag":41,"props":260,"children":261},{},[262],{"type":24,"value":263},"设计“预测-校正”迭代算法，利用弧长法追踪解路径，并结合SVD（奇异值分解）检测雅可比矩阵的奇异性来定位分岔点。",{"type":18,"tag":26,"props":265,"children":266},{"style":28},[267],{"type":18,"tag":31,"props":268,"children":270},{"src":269,"style":34,"alt":7},"/category/information/technology-blogs/banner/2026-1-15/5.jpg",[],{"type":18,"tag":37,"props":272,"children":273},{},[274],{"type":18,"tag":41,"props":275,"children":276},{},[277],{"type":24,"value":278},"步骤四： 多维可视化实现",{"type":18,"tag":37,"props":280,"children":281},{},[282],{"type":18,"tag":41,"props":283,"children":284},{},[285],{"type":24,"value":286},"实现了标准分岔图、能量演化图以及基于PCA降维的解路径轨迹演化图的绘制。",{"type":18,"tag":26,"props":288,"children":289},{"style":28},[290],{"type":18,"tag":31,"props":291,"children":293},{"src":292,"style":34,"alt":7},"/category/information/technology-blogs/banner/2026-1-15/6.jpg",[],{"type":18,"tag":106,"props":295,"children":297},{"id":296},"_3项目主要贡献",[298],{"type":24,"value":299},"3、项目主要贡献",{"type":18,"tag":55,"props":301,"children":302},{},[303,313,323,333],{"type":18,"tag":59,"props":304,"children":305},{},[306,308],{"type":24,"value":307},"**提出了对称性驱动的LQA改进算法：**",{"type":18,"tag":41,"props":309,"children":310},{},[311],{"type":24,"value":312},"成功将MindSpore Quantum应用于Z2格点规范场问题，通过对偶变换减少了变量数并降低了模型阶数。实验表明，改进后的方法相比朴素LQA算法，TTS（求解时间）提升了3倍以上，且在1000+节点规模下仍能在1秒内完成对称性计算。",{"type":18,"tag":59,"props":314,"children":315},{},[316,318],{"type":24,"value":317},"**构建了完整的解空间探索框架：**",{"type":18,"tag":41,"props":319,"children":320},{},[321],{"type":24,"value":322},"实现了基于同伦延拓的自动分岔点检测机制，能够在1小时内完成前5个分岔点的计算，解决了传统启发式算法易陷入局部最优的问题。",{"type":18,"tag":59,"props":324,"children":325},{},[326,328],{"type":24,"value":327},"**提供了丰富的可视化工具：**",{"type":18,"tag":41,"props":329,"children":330},{},[331],{"type":24,"value":332},"开发了PCA轨迹弧长演化图等工具，直观展示了高维能量景观中的解路径分岔行为。",{"type":18,"tag":59,"props":334,"children":335},{},[336,338],{"type":24,"value":337},"**交付了完整的API接口与性能基准测试：**",{"type":18,"tag":41,"props":339,"children":340},{},[341],{"type":24,"value":342},"为后续研究和应用提供了可靠的工具与评估标准。",{"type":18,"tag":37,"props":344,"children":345},{},[346],{"type":18,"tag":41,"props":347,"children":348},{},[349],{"type":24,"value":350},"# 03",{"type":18,"tag":37,"props":352,"children":353},{},[354],{"type":18,"tag":41,"props":355,"children":356},{},[357],{"type":24,"value":358},"攻坚纪实：当挑战化为突破",{"type":18,"tag":37,"props":360,"children":361},{},[362],{"type":18,"tag":41,"props":363,"children":364},{},[365],{"type":24,"value":366},"挑战一：启发式算法在受限空间中的“结构盲点”导致效率低下",{"type":18,"tag":55,"props":368,"children":369},{},[370,375,380],{"type":18,"tag":59,"props":371,"children":372},{},[373],{"type":24,"value":374},"问题根源：原始的Z2格点规范场问题包含海量的“非物理态”（不满足高斯定律约束的状态）。传统的模拟退火或朴素LQA算法不仅需要处理高阶相互作用，还会浪费大量计算资源在无效的非物理空间中搜索，导致收敛慢且难以找到全局最优。",{"type":18,"tag":59,"props":376,"children":377},{},[378],{"type":24,"value":379},"探索过程：问题的核心在于算法“不知道”物理系统的守恒律（对称性）。如果直接在高阶多项式上运行，计算开销呈指数级增长。那么解决问题的方法就主要两种：①用更高效的方法实现计算。②将问题转化为更小的/更有利于计算的形式。经过文献阅读并与导师讨论，后者具有更高的可行性。",{"type":18,"tag":59,"props":381,"children":382},{},[383],{"type":24,"value":384},"解决方案：引入基于图论的对称性计算。利用Kruskal算法构建最小生成森林来识别系统的循环基，以此定义对偶变换路径。通过将原问题映射为无约束的对偶自旋模型，直接剔除了所有非物理态，使搜索空间大幅缩减，从而显著提升了求解效率。",{"type":18,"tag":37,"props":386,"children":387},{},[388],{"type":18,"tag":41,"props":389,"children":390},{},[391],{"type":24,"value":392},"挑战二：复杂能量景观下的解路径追踪与分岔点定位",{"type":18,"tag":55,"props":394,"children":395},{},[396,401,406],{"type":18,"tag":59,"props":397,"children":398},{},[399],{"type":24,"value":400},"问题根源：在研究系统参数变化时，系统基态会发生定性突变（分岔）。普通的优化算法只能找到单一解，无法描绘出解随参数变化的完整路径，也难以检测到导致解结构改变的临界点。",{"type":18,"tag":59,"props":402,"children":403},{},[404],{"type":24,"value":405},"探索过程：尝试了传统的梯度下降/遗传算法，但无法处理多解共存和路径折返的情况。后与导师讨论，并调研了同伦方法，经过分析同伦方法的思想非常适合分岔点搜索，故基于同伦思想改进与实现分岔点搜索。",{"type":18,"tag":59,"props":407,"children":408},{},[409],{"type":24,"value":410},"解决方案：设计了同伦延拓分岔搜索算法。引入弧长法作为参数，使算法能追踪“回头”的折返路径。在每一步迭代中监测雅可比矩阵的特征值，利用SVD检测奇异点。一旦检测到分岔，自动派生新的求解器实例沿不同切线方向探索，最终成功绘制包含多个分支的全局分岔图。",{"type":18,"tag":37,"props":412,"children":413},{},[414],{"type":18,"tag":41,"props":415,"children":416},{},[417],{"type":24,"value":418},"# 04",{"type":18,"tag":37,"props":420,"children":421},{},[422],{"type":18,"tag":41,"props":423,"children":424},{},[425],{"type":24,"value":426},"开发者说：成长、思考与寄语",{"type":18,"tag":55,"props":428,"children":429},{},[430,440,450,460,470],{"type":18,"tag":59,"props":431,"children":432},{},[433,435,438],{"type":24,"value":434},"是什么机缘让你在开源之夏的诸多项目中选择了昇思MindSpore？在选择项目任务和撰写申请书的时候有哪些考虑和准备？",{"type":18,"tag":78,"props":436,"children":437},{},[],{"type":24,"value":439},"孙文杰：选择昇思MindSpore，主要是被其在科学计算与量子计算领域的深耕所吸引。它展现出的跨学科融合能力与我的研究兴趣高度契合，能在这个成熟的开源社区实战是非常难得的机会。基于这份认可，我在任务选择和申请书撰写上也做了充分的准备，首先是“扬长”，结合自己的技术栈锁定任务，因为只有在擅长且热爱的领域工作，才更容易产出高质量成果；其次是‘证能’，通过梳理过往的开源项目和问题分析过程，来证明自己既有开源情怀，又具备将项目落地的技术底气。",{"type":18,"tag":59,"props":441,"children":442},{},[443,445,448],{"type":24,"value":444},"此次开发工作与你以前的项目开发经历有何不可异同？",{"type":18,"tag":78,"props":446,"children":447},{},[],{"type":24,"value":449},"孙文杰：最大的不同是代码风格/规整性方面。过去的项目开发，比如科研过程中的代码编写，通常是基于自己风格的，而MindSpore对代码的规范性和简洁性会有更高的要求，让我学到了很多，也提高了自己代码的规范化水平。",{"type":18,"tag":59,"props":451,"children":452},{},[453,455,458],{"type":24,"value":454},"通过这个项目任务，你对开源有了什么更深刻的理解吗?",{"type":18,"tag":78,"props":456,"children":457},{},[],{"type":24,"value":459},"孙文杰：关于开源，我学到的一点是除了狭义的开源之外，“讨论”的过程更是非常重要的，基于昇思MindSpore提供的宝贵的平台，通过开源代码以及在项目中和大家讨论问题，能得到更多的收获，相互学习，共同成长与进步。",{"type":18,"tag":59,"props":461,"children":462},{},[463,465,468],{"type":24,"value":464},"作为学生参与开源项目，你认为最大的挑战是什么？又是如何克服的？",{"type":18,"tag":78,"props":466,"children":467},{},[],{"type":24,"value":469},"孙文杰：最大的挑战应该是在较为有限的时间内去学习一个与自己研究方向存在一定区别的方向，并且给出有效的改进方案。过程中，首先需要了解需要解决的问题（输入），需要达到的目标（输出），其中最重要的一点是，抓住问题主干，对问题构建自己的理解，并且时刻保持质疑和学习的态度。当有外界的输入与自己构建的问题的理解存在冲突的时候，不要全盘否定或者全盘接受，要分析和提出问题，并进行快速验证，最后改进和完善自己的“模型”。同时，对自己构建的问题模型，要保持“自洽”，因为那些系统中因为反馈产生越来越大的“振荡”（模拟信号）或者“非自洽”（数字信号，逻辑）的地方，往往就是问题的关键点。",{"type":18,"tag":59,"props":471,"children":472},{},[473,475,478],{"type":24,"value":474},"作为过来人，有没有什么话想对过去的自己/学弟学妹/刚加入昇思MindSpore的开发者说呢？",{"type":18,"tag":78,"props":476,"children":477},{},[],{"type":24,"value":479},"孙文杰：学而不思则罔，思而不学则殆。感谢昇思MindSpore给我提供了一个发挥自身价值，并在过程中不断思考、实践成长的机会，希望大家勇敢地申请和参与，通过在项目过程中广泛的交流和沟通与思考，相信大家都能学到很多，不断成长，变得更强！",{"title":7,"searchDepth":481,"depth":481,"links":482},4,[483,485,486,487,488,489,490],{"id":7,"depth":484,"text":7},3,{"id":111,"depth":484,"text":114},{"id":117,"depth":484,"text":123},{"id":186,"depth":484,"text":7},{"id":190,"depth":484,"text":193},{"id":196,"depth":484,"text":202},{"id":296,"depth":484,"text":299},"markdown","content:technology-blogs:zh:2026-1-15.md","content","technology-blogs/zh/2026-1-15.md","technology-blogs/zh/2026-1-15","md",1776506118342]