关于「指垫消退时机与纹型:一个不能唯一求解的反演问题」,《手纹的图灵斑图与拓扑》给出的定量结论是什么?

结论:正向映射把(峰值高度、消退速率、成嵴周数)三个参数压成一个组合量,纹型只读这个组合量。贝叶斯反演把指垫高度钉在 0.658 附近,成嵴周数却可以落在 10.05 到 13.25 之间,取决于假定的消退速率。十个手指的信息矩阵仍有一个精确为零的特征值。指纹是胚胎期的记录,但它记录的是一个乘积,而不是一份时间表。

数据口径:相学把它们当作写好的文书来读,本书则问三个可以计算的问题:这些纹怎样长出来?它们在人群中如何分布、彼此是否独立?从一枚成人指纹能否倒推胚胎期发生了什么?书中参数为示意,表格计数为合成数据,全部计算由书中程序生成;本书分析结构与概率,不对任何人的手作个人判断,也不是医疗、法律或经营建议。 第二章的二维有限差分模拟显示,一个系数的方向性就能把斑点变成平行嵴,细节点则由噪声决定。

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: 0.658, 10.05, 13.25

observed: whorl, ridge count 20
posterior onset week  11.26 +- 1.02, 90% 10.05 to 13.25
posterior rate r      0.221 +- 0.097, 90% 0.105 to 0.415
posterior height H    0.658 +- 0.057, 90% 0.571 to 0.757
correlation of tau and r in the posterior -0.780
if r were known exactly:
  r 0.15: onset week 11.72 +- 0.59
  r 0.30: onset week 10.42 +- 0.25
  r 0.45: onset week 10.11 +- 0.09
the print pins H; the week moves wherever r is assumed
查看完整验算程序
# Chapter 5. p52: Inverting a fingerprint into embryonic timing. Use the
# forward map of p51 and add a ridge count read from the print,
# RC ~ Normal(30 H, 2) (illustrative scale). Observed: a whorl with
# ridge count 20. Prior: onset week tau uniform on [10, 16], regression
# rate r uniform on [0.10, 0.50], asymmetry s uniform on [-0.5, 0.5].
# The posterior is computed exactly on a grid (no sampling).
import math

OBS_RC, SD = 20, 2.0
TAUS = [10 + 0.05 * i for i in range(121)]
RS = [0.10 + 0.005 * j for j in range(81)]
SS = [-0.5 + 0.05 * k for k in range(21)]


def like(tau, r):
    h = math.exp(-r * (tau - 9))
    if h < 0.55:
        return 0.0, h
    p_sym = len([s for s in SS if abs(s) < 0.3]) / len(SS)
    z = (OBS_RC - 30 * h) / SD
    return p_sym * math.exp(-z * z / 2), h


def moments(cells, k):
    w = math.fsum(c[0] for c in cells)
    m = math.fsum(c[0] * c[k] for c in cells) / w
    v = math.fsum(c[0] * (c[k] - m) ** 2 for c in cells) / w
    return m, math.sqrt(v)


def correlation(cells):
    w = math.fsum(c[0] for c in cells)
    (m1, s1), (m2, s2) = moments(cells, 1), moments(cells, 2)
    cov = math.fsum(c[0] * (c[1] - m1) * (c[2] - m2) for c in cells) / w
    return cov / (s1 * s2)


def interval(cells, k, q=0.05):
    w = math.fsum(c[0] for c in cells)
    pts = sorted((c[k], c[0] / w) for c in cells)
    acc, lo, hi = 0.0, None, None
    for x, p in pts:
        acc += p
        if lo is None and acc >= q:
            lo = x
        if hi is None and acc >= 1 - q:
            hi = x
    return lo, hi


if __name__ == "__main__":
    cells = []
    for tau in TAUS:
        for r in RS:
            l, h = like(tau, r)
            if l > 0:
                cells.append((l, tau, r, h))
    (mt, st), (mr, sr) = moments(cells, 1), moments(cells, 2)
    mh, sh = moments(cells, 3)
    rho = correlation(cells)
    print("observed: whorl, ridge count {}".format(OBS_RC))
    print("posterior onset week  {:.2f} +- {:.2f}, 90% {:.2f} to {:.2f}"
          .format(mt, st, *interval(cells, 1)))
    print("posterior rate r      {:.3f} +- {:.3f}, 90% {:.3f} to {:.3f}"
          .format(mr, sr, *interval(cells, 2)))
    print("posterior height H    {:.3f} +- {:.3f}, 90% {:.3f} to {:.3f}"
          .format(mh, sh, *interval(cells, 3)))
    print("correlation of tau and r in the posterior {:.3f}".format(rho))
    print("if r were known exactly:")
    for r0 in (0.15, 0.30, 0.45):
        sub = [c for c in cells if abs(c[2] - r0) < 1e-9]
        if not sub:
            print("  r {:.2f}: no onset week fits".format(r0))
            continue
        m, s = moments(sub, 1)
        print("  r {:.2f}: onset week {:.2f} +- {:.2f}".format(r0, m, s))
    print("the print pins H; the week moves wherever r is assumed")

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《手掌皮纹图灵反应扩散斑图、屈肌折痕拓扑分类与胚胎期应力反演》第 5 章「指垫消退时机与纹型:一个不能唯一求解的反演问题」(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-158 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026158,
  author    = {Fan, Yafu},
  title     = {Turing Reaction-Diffusion Patterns in Palm Dermatoglyphics, Flexion-Crease Topology and Embryonic Stress Inversion},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论