关于「季度母机收敛体检:流程、母机收敛判据计算器与预注册证伪方案」,《母机收敛判据》给出的定量结论是什么?

结论:第八章给出季度母机收敛体检、蒸馏小工具「母机收敛判据计算器」与预注册证伪方案,检验剪枝效果每组需审计 7,785 对规则。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。

数据口径:书中母机参数均为示意,不涉及任何服务器、域名、密钥或生产信息;全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营建议。 第四章用卡恩算法排序、用强连通分量查环修环,三百卷合成引用图中可比对为 18,828 对,占 0.420。

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: 7785

Test 1: audited pairs per arm (q00.010)
q10.008: 35,001 pairs per arm
q10.006: 7,785 pairs per arm
q10.005: 4,673 pairs per arm
Test 2: one-window power of t(b2) > 3 on case A
w end b2 E[t] power
814 0.792 0.97 0.021
1214 0.792 3.16 0.565
1224 0.792 1.62 0.085
1624 0.792 3.68 0.750
1640 0.792 1.98 0.153
2440 0.792 6.11 0.999
查看完整验算程序
# Chapter 8. p82: sample sizes for the pre-registered tests.
# Test 1 (pruning works): clash rate per audited pair among rules that
# passed screening (q1) against unscreened rules (q0); two-sided
# alpha 0.05, power 0.80, normal approximation for two proportions.
# Test 2 (quadratic law): power of the strict curvature alarm, t(b2)
# > 3, on a w-quarter window of case A means of p33 ending at quarter
# e, with Poisson noise sd = sqrt(mean). Expected t = b2 / se(b2);
# power = Phi(E[t] - 3) for one window.
# Rates and windows are illustrative.
import math
from statistics import NormalDist

Z = NormalDist()


def n_two_props(p0, p1, alpha=0.05, power=0.80):
    za = Z.inv_cdf(1 - alpha / 2)
    zb = Z.inv_cdf(power)
    pb = (p0 + p1) / 2
    num = (za * math.sqrt(2 * pb * (1 - pb))
           + zb * math.sqrt(p0 * (1 - p0) + p1 * (1 - p1))) ** 2
    return math.ceil(num / (p1 - p0) ** 2)


def case_a(quarters, a=24, eps=0.25, g=0.5, k=4):
    good, noise, mu = 0.0, 0.0, []
    for _ in range(quarters):
        good = good + (1 - eps) * a
        noise = noise + eps * a - min(k, g * noise)
        mu.append(0.02 * (noise * good + noise * (noise - 1) / 2)
                  + 0.0002 * good * (good - 1) / 2)
    return mu


def expected_t(ys):
    w = len(ys)
    s = [i - (w - 1) / 2 for i in range(w)]
    m2 = math.fsum(u * u for u in s) / w
    d = [u * u - m2 for u in s]
    sdd = math.fsum(x * x for x in d)
    b2 = math.fsum(x * y for x, y in zip(d, ys)) / sdd
    var = math.fsum(ys) / w
    return b2, b2 / math.sqrt(var / sdd)


if __name__ == "__main__":
    print("Test 1: audited pairs per arm (q0 = 0.010)")
    for q1 in (0.008, 0.006, 0.005):
        print("  q1 = {:.3f}: {:>7,} pairs per arm".format(
            q1, n_two_props(0.010, q1)))
    print("Test 2: one-window power of t(b2) > 3 on case A")
    print("   w   end    b2   E[t]   power")
    mu = case_a(40)
    for w, e in ((8, 14), (12, 14), (12, 24), (16, 24), (16, 40),
                 (24, 40)):
        b2, t = expected_t(mu[e - w:e])
        print("  {:>2}  {:>4}  {:.3f}  {:>5.2f}  {:.3f}".format(
            w, e, b2, t, Z.cdf(t - 3.0)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《三百卷定理依赖图拓扑排序与母机收敛判据、立项噪音与剪枝增益的闭环停机条件母典》第 8 章「季度母机收敛体检:流程、母机收敛判据计算器与预注册证伪方案」(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-332 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026332,
  author    = {Fan, Yafu},
  title     = {Topological Ordering of the Theorem-Dependency Graph of Three Hundred Volumes and a Convergence Criterion for the Self-Evolving Engine: Halting Conditions between Proposal Noise and Pruning Gain},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论