关于「最大似然整合:多通道带来的是精度提升,不是乘法放大」,《三秒薄切片信任》给出的定量结论是什么?

结论:第二章说明三路等精度时标准差最多缩到 0.577,三路判别力合并为 1.044,只是最强单路的 1.305 倍。

数据口径:书中参数为示意;全部计算由书中程序生成;本书只写如何让真实、可兑现的信号在前三秒清晰一致,不写任何伪造资质或假权威的手法,不宣称任何气味的健康功效,是非个人化经营建议。 第一章区分信度与效度,示意参数下三秒判断只达到效度上限的 0.612,与长期评价的一致率为 0.579。

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: 1.044, 1.305

A. single channels: d' V/A/O0.8, 0.6, 0.3
d' 0.8: accuracy 0.655
d' 0.6: accuracy 0.618
d' 0.3: accuracy 0.560
B. three channels, equal noise correlation rho
rho d'_comb /best accuracy
0.01.044 1.305 0.699
0.30.885 1.106 0.671
0.60.869 1.086 0.668
C. slogans vs the bound (rho0)
naive sum of d'1.700
product of d'0.144
optimal1.044
D. two equal channels d' 0.8: combined 1.1310.8 x 1.414
查看完整验算程序
# Chapter 2. p22: how much three channels add to telling a reliable seller
# from an unreliable one. Each channel separates the two kinds by d'_i in
# noise units; with noise correlation matrix S the optimal combined d' is
# sqrt(d' S^-1 d'). Independent noise gives sqrt(sum of d'_i^2): squares add,
# the d' values do not multiply. Accuracy of an unbiased two-way call is
# Phi(d'/2). The d' values and correlations are illustrative.
import math

D = (0.8, 0.6, 0.3)


def phi(x):
    return 0.5 * (1.0 + math.erf(x / math.sqrt(2.0)))


def solve(a, b):
    n = len(b)
    m = [list(a[i]) + [b[i]] for i in range(n)]
    for c in range(n):
        p = max(range(c, n), key=lambda r: abs(m[r][c]))
        m[c], m[p] = m[p], m[c]
        for r in range(n):
            if r != c:
                f = m[r][c] / m[c][c]
                m[r] = [x - f * y for x, y in zip(m[r], m[c])]
    return [m[i][n] / m[i][i] for i in range(n)]


def combined(d, rho):
    n = len(d)
    s = [[1.0 if i == j else rho for j in range(n)] for i in range(n)]
    x = solve(s, list(d))
    return math.sqrt(math.fsum(a * b for a, b in zip(d, x)))


if __name__ == "__main__":
    print("A. single channels: d' V/A/O = {}".format(
        ", ".join("{:.1f}".format(x) for x in D)))
    for x in D:
        print("   d' {:.1f}: accuracy {:.3f}".format(x, phi(x / 2.0)))
    print("B. three channels, equal noise correlation rho")
    print("   rho   d'_comb  /best   accuracy")
    for rho in (0.0, 0.3, 0.6):
        dc = combined(D, rho)
        print("   {:.1f}   {:.3f}    {:.3f}   {:.3f}".format(
            rho, dc, dc / max(D), phi(dc / 2.0)))
    print("C. slogans vs the bound (rho = 0)")
    print("   naive sum of d'  {:.3f}".format(math.fsum(D)))
    print("   product of d'    {:.3f}".format(D[0] * D[1] * D[2]))
    print("   optimal          {:.3f}".format(combined(D, 0.0)))
    two = combined((0.8, 0.8), 0.0)
    print("D. two equal channels d' 0.8: combined {:.3f} = 0.8 x {:.3f}"
          .format(two, two / 0.8))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《三秒薄切片信任捕获:视听嗅三通道同频一致性与瞬间权威建立母典》第 2 章「最大似然整合:多通道带来的是精度提升,不是乘法放大」(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-419 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026419,
  author    = {Fan, Yafu},
  title     = {Three-Second Thin-Slicing Trust: Visual-Auditory-Olfactory Congruence and Instant Authority Signalling},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论