关于「结构同构:谱指纹、穷举映射与组合典型性」,《量子启发与猜想证伪》给出的定量结论是什么?

结论:结构相似与字面相似是两回事。热传导与电路的奇异值完全相同,谱距离 0.0000,词表却毫不重合;厨房场景用词重合 0.308,谱距离却有 1.6179。穷举映射确认热传导到电路 6 条关系全部对齐,到厨房只对上 2 条,且有 90 个映射并列。谱相同不等于同构,星形图与四边形加孤立点就是反例。Uzzi 的典型性 z 分数能区分常规与新颖,但它的门槛依赖语料构成。

数据口径:书中参数为示意,全部计算由书中程序生成。 本书不构成投资、学术评审或量子硬件方面的建议。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 0.0000, 0.308, 1.6179

heat top singular values2.2361 1.4142 1.4142
electric top singular values2.2361 1.4142 1.4142
cooking top singular values2.8531 0.9273 0.0000
spectral distance heat vs electric0.0000
spectral distance heat vs cooking1.6179
word overlap heat vs electric: Jaccard 0.000
word overlap heat vs cooking: Jaccard 0.308
查看完整验算程序
# Chapter 5. p51: singular values as a structural fingerprint.
# Three illustrative concept-relation matrices (6 concepts each, entry
# A[i][j] = strength of "i drives j"): heat conduction, an RC electrical
# circuit with the concepts listed in a different order, and a cooking
# scene that shares the heat vocabulary but not the relations. Top
# singular values come from power iteration on A^T A with deflation, in
# pure Python. Relabelling concepts permutes rows and columns and leaves
# the singular values unchanged; shared words do not.
import math

HEAT = ["hot body", "cold body", "temperature gap", "heat flux",
        "conductivity", "heat capacity"]
HEAT_E = [(0, 2, 1), (1, 2, 1), (2, 3, 2), (4, 3, 1), (3, 1, 1), (5, 1, 1)]
ELEC = ["current", "capacitance", "high node", "conductance",
        "low node", "voltage drop"]
ELEC_E = [(2, 5, 1), (4, 5, 1), (5, 0, 2), (3, 0, 1), (0, 4, 1), (1, 4, 1)]
COOK = ["hot pan", "cold food", "temperature", "heat", "oil", "lid"]
COOK_E = [(0, 1, 1), (0, 2, 1), (0, 3, 2), (0, 4, 1), (0, 5, 1), (3, 2, 1)]


def matrix(edges, n=6):
    a = [[0.0] * n for _ in range(n)]
    for i, j, w in edges:
        a[i][j] = float(w)
    return a


def gram(a):
    n = len(a)
    return [[math.fsum(a[k][i] * a[k][j] for k in range(n))
             for j in range(n)] for i in range(n)]


def top_singular(a, count=3, iters=500):
    g = gram(a)
    n = len(g)
    out = []
    for r in range(count):
        v = [1.0 + 0.1 * ((i * 7 + r) % 5) for i in range(n)]
        lam = 0.0
        for _ in range(iters):
            w = [math.fsum(g[i][j] * v[j] for j in range(n))
                 for i in range(n)]
            norm = math.sqrt(math.fsum(x * x for x in w))
            if norm < 1e-14:
                break
            v = [x / norm for x in w]
            lam = norm
        out.append(math.sqrt(max(lam, 0.0)))
        g = [[g[i][j] - lam * v[i] * v[j] for j in range(n)]
             for i in range(n)]
    return out


if __name__ == "__main__":
    spec = {}
    for name, edges in (("heat", HEAT_E), ("electric", ELEC_E),
                        ("cooking", COOK_E)):
        spec[name] = top_singular(matrix(edges))
        print("  {:<8} top singular values: {}".format(
            name, "  ".join("{:.4f}".format(x) for x in spec[name])))
    for a, b in (("heat", "electric"), ("heat", "cooking")):
        d = math.sqrt(math.fsum((x - y) ** 2
                                for x, y in zip(spec[a], spec[b])))
        print("  spectral distance {} vs {}: {:.4f}".format(a, b, d))
    shared = set(w for c in HEAT for w in c.split())
    for name, words in (("electric", ELEC), ("cooking", COOK)):
        other = set(w for c in words for w in c.split())
        jac = len(shared & other) / len(shared | other)
        print("  word overlap heat vs {}: Jaccard {:.3f}".format(name, jac))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《量子启发式组合寻优、拉马努金式直觉猜想与突破性著作推演母典》第 5 章「结构同构:谱指纹、穷举映射与组合典型性」(版本 1.0.0). 5D Library. DOI 待发布

DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-364 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026364,
  author    = {Fan, Yafu},
  title     = {Quantum-Inspired Combinatorial Search, Abductive Intuition and Computationally Verified Breakthrough Monograph Engineering},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论