现代 qLDPC 构造
现代 qLDPC 构造的比较与导航聚合主题;按代数输入、商/提升/扭曲机制、已证参数、有限实例和执行证据连接各独立叶子码族。
- 成熟度
- 核心基础
- 重要度
- 核心主题
- 证据
- 4 篇代表来源
主题要点
本页只给横向选型维度、构造包含关系和证据边界,不复述 BP、LP、Fiber-bundle、2BGA、GB、BB、Tanner 或 Expander 叶子的完整定义。
全栈位置与直接关系
code/code/advanced-qldpc现代 qLDPC 构造的比较与导航聚合主题;按代数输入、商/提升/扭曲机制、已证参数、有限实例和执行证据连接各独立叶子码族。
本页只给横向选型维度、构造包含关系和证据边界,不复述 BP、LP、Fiber-bundle、2BGA、GB、BB、Tanner 或 Expander 叶子的完整定义。
code/code/advanced-qldpcCode dossier
这里汇总参数景观中的结构化记录、码族谱系中的构造说明、执行路线与外部知识库,作为完整主题页的代码专属延伸。
聚合主题无单一 [[n,k,d]];有限坐标与阈值必须回到叶子和实例。
按 HGP/BP/LP/Fiber/2BGA/GB/BB/Tanner/Expander 独立叶子组织的横向比较。
按 HGP/BP/LP/Fiber/2BGA/GB/BB/Tanner/Expander 独立叶子组织的横向比较。
balanced product
lifted product
asymptotically good qLDPC
quantum Tanner / expander-based
Advanced qLDPC Comparative Workflow:叶子构造选择、有限实例门槛、统一矩阵/decoder 契约、证据归一化与跨族比较。当前路线绑定 Select decoder by leaf family、Cross-family decoder audit,并覆盖 Select finite leaf instance、Normalize syndrome contract。
聚合主题无单一 [[n,k,d]];有限坐标与阈值必须回到叶子和实例。
Deep reference
定义、代数构造、保护能力与执行证据均按来源段落独立维护;中文稿经过术语整理,英文稿保留用于逐段对照。
本页是现代 qLDPC 构造的导航与比较层,不代表一个单独的校验矩阵族。它回答的是‘输入对象是什么、用了哪一种乘积/提升/商/扭曲、定理保证到哪里、有没有可复现有限实例和执行证据’,而不是把所有高率稀疏码压成一组 [[n,k,d]]。具体定义、公式与论文结论由叶子主题承担。
This page is a navigation and comparison layer for modern qLDPC constructions, not a single check-matrix family. It organizes input objects, product/lift/quotient/twist mechanisms, proved guarantees, reproducible finite instances, and execution evidence. Definitions and theorems remain on the leaf topics.
构造选择首先看代数输入:HGP 使用经典校验矩阵;homological product 使用链复形;balanced product 使用带共同群作用的复形并取商;lifted product 使用环或群代数矩阵;fiber bundle 增加沿底空间的扭曲;2BGA 把 LP 缩到一乘一群代数元素;GB 与 BB 再分别选择循环群与二循环 Abelian 群。
| 叶子族 | 最少可复现数据 |
|---|---|
| HGP / Expander | 两个 seed 矩阵或二分图 |
| BP / Fiber | 群作用、底/纤维复形、轨道或连接 |
| LP / 2BGA | 系数环/群、基矩阵或群代数元素 |
| GB / BB | 循环长度与生成多项式 |
| Tanner | Cayley 复形、局部码与粘合规则 |
Construction choice begins with the algebraic input: HGP uses classical checks; homological products use chain complexes; BP quotients complexes with a shared group action; LP uses matrices over rings or group algebras; fiber bundles add twisting; 2BGA reduces LP to one-by-one group-algebra inputs; GB and BB then select cyclic and two-factor Abelian groups.
| Leaf family | Minimum reproducibility data |
|---|---|
| HGP / Expander | two seed matrices or bipartite graphs |
| BP / Fiber | group actions, base/fiber complexes, orbits or connection |
| LP / 2BGA | coefficient ring/group and base matrix or group elements |
| GB / BB | cycle lengths and generating polynomials |
| Tanner | Cayley complex, local codes, and gluing rules |
比较保护能力时必须把码率、距离、校验权重、量子比特度和综合征冗余分开。‘qLDPC’只约束局部稀疏性;常数率、线性或近线性距离、single-shot 综合征保护以及有效 circuit distance 都需要额外定理或线路证据。聚合页不把某个子族的优势传播给其他叶子。
Rate, distance, check weight, qubit degree, and syndrome redundancy are separate axes. The qLDPC label constrains sparsity only; constant rate, linear or almost-linear distance, single-shot syndrome protection, and effective circuit distance require additional theorems or circuit evidence.
只能在统一渐近变量与假设后横向比较:
逻辑门能力不是 qLDPC 聚合属性。可用机制包括链映射、群自同构诱导的置换、fold-transversal Clifford、码切换、逻辑 Pauli 乘积测量和门传送;每条结论都必须绑定具体有限码、逻辑基、辅助块、测量次序和故障传播分析。
Logical-gate capability is not an aggregate qLDPC property. Chain maps, group-automorphism permutations, fold-transversal Cliffords, code switching, logical-Pauli measurements, and gate teleportation all require a finite code, logical basis, ancilla blocks, measurement order, and fault-propagation analysis.
解码器按假设选型:Expander/Tanner 的局部算法依赖扩展与鲁棒性;一般稀疏 CSS 矩阵常以 BP+OSD 作有限长度基线;GB/BB 还可利用循环与综合征码结构;擦除、偏置、退极化、phenomenological 与 circuit-level 噪声不能共用一条阈值。聚合页只给入口,不保存脱离实例的‘最佳解码器’。
Decoder choice follows the hypotheses: local expander/Tanner algorithms need expansion and robustness; generic sparse CSS matrices often use BP+OSD as a finite-length baseline; GB/BB can exploit cyclic and syndrome-code structure. Erasure, biased, depolarizing, phenomenological, and circuit-level thresholds are not interchangeable.
常数校验权重不自动产生低深度或容错综合征线路。执行层还要登记 ancilla 数、调度冲突、连接图、远程纠缠、hook error、测量重复、实时解码时限与有效 circuit distance。没有这些字段时,页面只能声称静态码构造,不声称可执行容错存储器。
Constant check weight does not automatically give a shallow or fault-tolerant syndrome circuit. Execution evidence must record ancillas, scheduling conflicts, connectivity, remote entanglement, hook errors, repeated measurement, real-time decoding, and effective circuit distance.
入库时先选择记录层级:渐近族保存定理假设与标度;有限实例保存完整构造种子、展开矩阵和逻辑算子;执行工件再保存线路、噪声与解码配置。只含 family 名或 [[n,k,d]] 的条目不足以建立 QECDB/QECirc 精确身份。
Ingestion begins by selecting the record level. Asymptotic families store hypotheses and scaling; finite instances store complete seeds, expanded matrices, and logical operators; execution artifacts add circuits, noise, and decoder configuration. A family label or [[n,k,d]] alone is insufficient for strict QECDB/QECirc identity.
本页采用‘构造包含’而非‘名称相似’组织谱系:HGP 是 homological product 特例;BP、LP 与 fiber-bundle 在特定表示下可以互相改写;2BGA 是一乘一群代数 LP,也可表为 BP;GB 是循环群 2BGA;BB 是二循环 Abelian 2BGA。构造层可改写不等于任意有限条目已经通过置换、缩放或局部 Clifford 等价检验。
The hierarchy follows construction containment: HGP specializes homological products; BP, LP, and fiber bundles admit reformulations under stated data; 2BGA is a one-by-one group-algebra LP and also has a BP formulation; GB is cyclic 2BGA; BB is two-factor Abelian 2BGA. Construction-level reformulation is not a finite permutation, scaling, or local-Clifford equivalence certificate.
查表顺序建议固定为:先看记录层级,再看代数输入和等价关系;随后核对定理量词、有限矩阵身份、噪声模型与解码器;最后才比较阈值和资源。若链接只能落到数据库首页或更宽的父类筛选,本页显示‘尚无精确对应条目’,不制造有值但无意义的导航。
Lookup order should be: record level; algebraic input and equivalence relation; theorem quantifiers; finite matrix identity; noise model and decoder; only then thresholds and resources. A database home page or broader parent-family filter remains unavailable rather than becoming a misleading link.
Cross-family comparison requires a common asymptotic variable and explicit hypotheses. Quantum expanders have constant rate and square-root distance; fiber bundles prove rate and distance ; explicit distance-balanced BP gives , ; LP and quantum Tanner families reach constant rate with almost-linear or linear distance under their respective hypotheses. 2BGA, GB, and BB parameters remain group- and matrix-specific.