关于「衍生拓扑:从母书编译,还是从上一层编译」,《衍生品的出处与保真》给出的定量结论是什么?
结论:第五章比较衍生拓扑:不做检查时链形五件产品合计错 28.07 条事实,星形错 10.00 条;把编译成本与错误代价放进同一个式子,成本最低的通常是中枢形。
数据口径:书中参数为示意,代价用记账点数而非价格;全部计算由书中程序生成。 书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。
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: 10.00, 28.07
| unchecked, per-step fidelity 0.950 | |||||||
| topology | summary | script | slides | quiz | clip | mean | wrong |
| chain | 0.9500 | 0.9025 | 0.8574 | 0.8145 | 0.7738 | 0.8596 | 28.07 |
| star | 0.9500 | 0.9500 | 0.9500 | 0.9500 | 0.9500 | 0.9500 | 10.00 |
| re-anchor | 0.9500 | 0.9025 | 0.9500 | 0.9025 | 0.9500 | 0.9310 | 13.80 |
| with master check D=0.9, per-step fidelity 0.995 | |||||||
| topology | summary | script | slides | quiz | clip | mean | wrong |
| chain | 0.9950 | 0.9900 | 0.9851 | 0.9801 | 0.9752 | 0.9851 | 2.98 |
| star | 0.9950 | 0.9950 | 0.9950 | 0.9950 | 0.9950 | 0.9950 | 1.00 |
| re-anchor | 0.9950 | 0.9900 | 0.9950 | 0.9900 | 0.9950 | 0.9930 | 1.40 |
查看完整验算程序
# Chapter 5. p51: compile every product from the master, not in a chain.
# Five products: summary, script, slides, quiz, clip. In a chain each is
# compiled from the one before; in a star each is compiled from the
# master; "re-anchor" compiles every second product from the master.
# Per-compilation fidelity p = 0.95; with a check of detection rate
# D = 0.9 against the master and repair, the per-step loss becomes
# (1 - p)(1 - D). 40 facts. Illustrative parameters.
import math
N, P, D = 40, 0.95, 0.9
NAMES = ["summary", "script", "slides", "quiz", "clip"]
TOPO = {
"chain": [0, 1, 2, 3, 4], # parent index (0 = master)
"star": [0, 0, 0, 0, 0],
"re-anchor": [0, 1, 0, 3, 0],
}
def depths(par):
d = []
for j, p in enumerate(par):
d.append(1 if p == 0 else d[p - 1] + 1)
return d
for checked in (False, True):
q = 1 - (1 - P) * (1 - D) if checked else P
tag = "with master check D=%.1f" % D if checked else "unchecked"
print("%s, per-step fidelity %.3f" % (tag, q))
print("%-10s" % "topology" + "".join("%8s" % n for n in NAMES)
+ "%8s %7s" % ("mean", "wrong"))
for name, par in TOPO.items():
f = [q ** d for d in depths(par)]
mean = math.fsum(f) / len(f)
wrong = math.fsum(N * (1 - x) for x in f)
print("%-10s" % name + "".join("%8.4f" % x for x in f)
+ "%8.4f %7.2f" % (mean, wrong))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《母书多模态衍生编译与版权链、单部专著到播客投影片题库的结构同态流水线母典》第 5 章「衍生拓扑:从母书编译,还是从上一层编译」(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-268 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026268,
author = {Fan, Yafu},
title = {Multimodal Derivative Compilation and Provenance Chains: A Structure-Preserving Pipeline from One Monograph to Podcasts, Slides and Quiz Banks},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}