关于「邻近可能图谱:一步可达的组合与陌生度的取舍」,《存量资产的超模串联》给出的定量结论是什么?
结论:B 段把十次试验排成一个季度的组合。全部用熟悉组合(\(u = 0.1\)),总价值均值 11.77,至少一次突破的概率只有 0.0057,最差一成的总价值是 9.49。放进 1 次陌生试验(\(u = 0.8\)),突破概率升到 0.0390,最差一成降到 9.16;放进 2 次,突破概率 0.0712,最差一成 8.97;放进 3 次,突破概率 0.1023,最差一成 8.72。十次全部陌生,突破概率 0.2927,最差一成跌到 7.22。
数据口径:在示意模型里,六项资产两两配对的十五对中有十二对互补、三对替代,全部打开时月度价值 23,104,各自相加只有 13,668。 书中参数为示意,不对应任何真实业务、客户、达人或成交;全部计算由书中程序生成。
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: 0.0057, 0.0390, 0.0712, 0.1023, 0.2927, 0.8
| A. one trial by unfamiliarity u | |
| u mu(u) sigma mean value median P(value > 5) | |
| 0.00 | 0.000 0.40 1.083 1.000 0.0000 |
| 0.10 | 0.048 0.48 1.177 1.049 0.0006 |
| 0.25 | 0.075 0.60 1.290 1.078 0.0053 |
| 0.40 | 0.048 0.72 1.360 1.049 0.0151 |
| 0.60 -0.072 | 0.88 1.371 0.931 0.0280 |
| 0.80 -0.288 | 1.04 1.288 0.750 0.0340 |
| 1.00 -0.600 | 1.20 1.127 0.549 0.0328 |
| B. ten trials a quarter, k unusual (u | 0.8) |
| k mean total P(> | 1 hit) P10 of total |
| 0 | 11.77 0.0057 9.49 |
| 1 | 11.88 0.0390 9.16 |
| 2 | 11.99 0.0712 8.97 |
| 3 | 12.10 0.1023 8.72 |
| 5 | 12.32 0.1614 8.24 |
| 10 | 12.88 0.2927 7.22 |
查看完整验算程序
# Chapter 3. p33: unfamiliarity, mean and spread of a recombination.
# Illustrative shape, not a fitted curve: the log value of one chaining
# trial with unfamiliarity u in [0, 1] is normal with
# mu(u) = 0.6 u - 1.2 u^2 and sigma(u) = 0.4 + 0.8 u,
# so the mean of the log rises then falls while the spread keeps rising
# (Fleming 2001 reports lower average usefulness and higher variability
# for new combinations). Value 1 = a typical familiar trial; a
# "breakthrough" is value > 5.
# B. A quarter of ten trials: k unusual trials (u = 0.8) and 10 - k
# conventional ones (u = 0.1). Mean total, chance of at least one
# breakthrough, and the 10th percentile of the total (Monte Carlo).
import math
import random
HIT = 5.0
def mu(u):
return 0.6 * u - 1.2 * u * u
def sd(u):
return 0.4 + 0.8 * u
def mean_value(u):
return math.exp(mu(u) + sd(u) ** 2 / 2.0)
def p_hit(u):
z = (math.log(HIT) - mu(u)) / sd(u)
return 0.5 * math.erfc(z / math.sqrt(2.0))
if __name__ == "__main__":
print("A. one trial by unfamiliarity u")
print(" u mu(u) sigma mean value median P(value > 5)")
for u in (0.0, 0.1, 0.25, 0.4, 0.6, 0.8, 1.0):
print(" {:.2f} {:>6.3f} {:>6.2f} {:>11.3f} {:>8.3f} {:>13.4f}"
.format(u, mu(u), sd(u), mean_value(u), math.exp(mu(u)),
p_hit(u)))
print("B. ten trials a quarter, k unusual (u = 0.8)")
print(" k mean total P(>= 1 hit) P10 of total")
rng = random.Random(37503)
for k in (0, 1, 2, 3, 5, 10):
mean_t = math.fsum([mean_value(0.8)] * k +
[mean_value(0.1)] * (10 - k))
p_none = (1 - p_hit(0.8)) ** k * (1 - p_hit(0.1)) ** (10 - k)
tot = []
for _ in range(20000):
vals = [math.exp(rng.gauss(mu(0.8), sd(0.8))) for _ in range(k)]
vals += [math.exp(rng.gauss(mu(0.1), sd(0.1)))
for _ in range(10 - k)]
tot.append(math.fsum(vals))
tot.sort()
print(" {:>2} {:>11.2f} {:>13.4f} {:>14.2f}".format(
k, mean_t, 1 - p_none, tot[len(tot) // 10]))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《存量武器库超模组合代数、邻近可能联想图谱与跨业务火力串联母典》第 3 章「邻近可能图谱:一步可达的组合与陌生度的取舍」(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-375 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026375,
author = {Fan, Yafu},
title = {Supermodular Weapon-Arsenals, Adjacent-Possible Recombination Graphs and Cross-Asset Cashflow Chaining},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}