关于「稀疏的公理网络:度分布、聚类、连通块与模块度」,《四百卷自持回路总纲》给出的定量结论是什么?
结论:把新词率降到 0.800,差别开始出现:累积优势下不同词降到 1,064.4 个,最常用词出现 9.4 次,而有 8.6 个词出现五次以上。降到 0.500,累积优势给出明显的枢纽,最常用词平均出现 58.2 次,五次以上的词有 44.9 个;均匀复用下最常用词只出现 10.7 次。西蒙模型的尾部指数也说明同一件事:新词率 0.933 时指数是 15.93,尾部陡到几乎没有尾;0.800 时是 6.00;0.500 时才降到 3.00,接近引文网络常见的量级。
数据口径:书库是群岛,不是全息网;合成模型显示,半数新卷各做一次桥接引用,就能让大部分新卷接入同一巨连通块。 示意基准下 g 为 0.018,临界规模 55.6,再投入率可行带 0.294 到 0.529;相图里同时满足增长与个人报酬条件的格子只占 0.231。
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: 0.500, 0.800, 0.933, 10.7, 1064.4, 15.93
| real: distinct 1,241 hist [1167, 64, 6, 3, 1] | |
| p_new mode distinct once twice | 3x 4x 5+ top |
| 0.933 pref | 1,241.1 1,163 68.4 7.7 1.2 0.3 4.1 |
| 0.933 unif | 1,240.0 1,156 77.8 5.4 0.4 0.0 3.4 |
| 0.800 pref | 1,064.4 887 126.8 31.4 10.6 8.6 9.4 |
| 0.800 unif | 1,064.4 851 171.5 33.7 6.7 1.6 5.2 |
| 0.500 pref | 665.6 445 109.6 44.6 21.8 44.9 58.2 |
| 0.500 unif | 663.9 331 165.4 83.7 41.5 41.8 10.7 |
| Simon tail exponent at p_new 0.933 | 15.93 |
| Simon tail exponent at p_new 0.800 | 6.00 |
| Simon tail exponent at p_new 0.500 | 3.00 |
查看完整验算程序
# Chapter 2. p22: cumulative advantage in keyword reuse (Simon / Price)
# A stream of 1,330 keyword tokens (the size of the real B-001..B-332
# list). Each token is a new keyword with probability p_new; otherwise
# it reuses an old one, chosen in proportion to past use (cumulative
# advantage) or uniformly (no advantage). Real count-of-counts from the
# public index: 1,167 keywords used once, 64 twice, 6 three times,
# 3 four times, 1 five times. Fixed seed, 200 replicates per setting.
import math
import random
TOKENS = 1330
REAL = [1167, 64, 6, 3, 1] # used 1, 2, 3, 4, 5+ times
rng = random.Random(40822)
def stream(p_new, pref):
toks, counts = [], []
for _ in range(TOKENS):
if not toks or rng.random() < p_new:
counts.append(1)
toks.append(len(counts) - 1)
else:
if pref:
j = toks[rng.randrange(len(toks))]
else:
j = rng.randrange(len(counts))
counts[j] += 1
toks.append(j)
return counts
def profile(p_new, pref, reps=200):
hist = [0.0] * 5
tops = []
for _ in range(reps):
c = stream(p_new, pref)
for v in c:
hist[min(v, 5) - 1] += 1
tops.append(max(c))
return [h / reps for h in hist], math.fsum(tops) / reps
def main():
real_d = int(math.fsum(REAL))
print("real: distinct {:,} hist {}".format(real_d, REAL))
print("p_new mode distinct once twice 3x 4x 5+ top")
for p_new in (0.933, 0.80, 0.50):
for pref in (True, False):
h, top = profile(p_new, pref)
d = math.fsum(h)
print("{:.3f} {:>5} {:>8,.1f} {:>5,.0f} {:>6.1f} {:>4.1f}"
" {:>4.1f} {:>4.1f} {:>4.1f}".format(
p_new, "pref" if pref else "unif", d, h[0], h[1],
h[2], h[3], h[4], top))
for p_new in (0.933, 0.80, 0.50):
print("Simon tail exponent at p_new {:.3f}: {:.2f}".format(
p_new, 1 + 1 / (1 - p_new)))
main()唯一数据源与引用
范雅福(Fan, Y.). (2026). 《四百卷全息知识共同体、现金流与算力自持闭环及无限创作自由态母典》第 2 章「稀疏的公理网络:度分布、聚类、连通块与模块度」(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-408 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026408,
author = {Fan, Yafu},
title = {The Four-Hundred-Volume Holographic Knowledge Commonwealth: Self-Sustaining Cashflow-Compute Loops and Unbounded Creative Sovereignty},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}