关于「视觉卡片压缩:投影片每页信息量上限、按依赖切页的动态规划与卡片字数」,《专著改编的信息保留代数》给出的定量结论是什么?
结论:第三章给出每页 3 条要点时每分钟留住 4.0909 条,平铺排版的卡片在 40 字时带走最多,为 0.7113。
数据口径:书中参数为示意;全部计算由书中程序生成;本书不是法律意见,也不是个人化的经营或投资建议。 第六章在 40 个制作工时下求出播客、投影片与题库的最优组合,价值 2.0685,不做题库只有 1.4164;事实核对应放在上游;每周 2 次提醒使期望学习次数最高,为 6.2792,过量提醒有害,对脆弱学习者阈值更低;只改编有权改编的书,合成内容必须标识。
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: 0.7113, 40
| chars | on card | front-loaded | evenly spread | read to the end |
|---|---|---|---|---|
| 10 | 0.3297 | 0.3296 | 0.3296 | 0.9972 |
| 20 | 0.5507 | 0.5460 | 0.5443 | 0.9400 |
| 30 | 0.6988 | 0.6761 | 0.6641 | 0.7988 |
| 40 | 0.7981 | 0.7476 | 0.7113 | 0.6276 |
| 50 | 0.8647 | 0.7843 | 0.7111 | 0.4717 |
| 60 | 0.9093 | 0.8026 | 0.6845 | 0.3436 |
| 80 | 0.9592 | 0.8158 | 0.6023 | 0.1797 |
| 100 | 0.9817 | 0.8188 | 0.5187 | 0.0914 |
| 140 | 0.9963 | 0.8197 | 0.3908 | 0.0262 |
| median characters read in one look: 48.1 | ||||
| best front-loaded: 140 characters, 0.8197 taken away | ||||
| best evenly spread: 40 characters, 0.7113 taken away |
查看完整验算程序
# Chapter 3. p33: how many characters a visual card should carry.
# A card compressed from a passage keeps a share f(W) = 1 - exp(-W / 25)
# of the passage's information when it carries W Chinese characters (a
# saturating coverage curve of the kind p11 produces). A viewer looks at
# a card for D seconds, lognormal with median 6 s and log-sd 0.5, and
# reads at R characters per second, lognormal with median 8 and log-sd
# 0.25, so min(W, D * R) characters get read. Two layouts: front-loaded
# (most important first, the viewer takes f of what was read) and
# evenly spread (the viewer takes the read share of f(W)).
# Fixed-seed Monte Carlo, 40,000 viewers. Every parameter illustrative.
import math
import random
rng = random.Random(20261112)
VIEWERS = 40000
READ = [math.exp(math.log(6.0) + 0.5 * rng.gauss(0.0, 1.0))
* math.exp(math.log(8.0) + 0.25 * rng.gauss(0.0, 1.0))
for _ in range(VIEWERS)]
def f(w):
return 1.0 - math.exp(-w / 25.0)
print("chars on card front-loaded evenly spread read to the end")
best = {"front": (0.0, 0), "even": (0.0, 0)}
for w in (10, 20, 30, 40, 50, 60, 80, 100, 140):
front = math.fsum(f(min(w, r)) for r in READ) / VIEWERS
even = math.fsum(f(w) * min(1.0, r / w) for r in READ) / VIEWERS
done = len([r for r in READ if r >= w]) / VIEWERS
best["front"] = max(best["front"], (front, w))
best["even"] = max(best["even"], (even, w))
print("{:5d} {:7.4f} {:12.4f} {:13.4f} {:15.4f}".format(
w, f(w), front, even, done))
print("median characters read in one look: {:.1f}".format(
sorted(READ)[VIEWERS // 2]))
for k, label in (("front", "front-loaded"), ("even", "evenly spread")):
print("best {}: {} characters, {:.4f} taken away".format(
label, best[k][1], best[k][0]))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《一书多吃内容裂变代数:从硬核专著自动编译双人播客、导学投影片与通关题库》第 3 章「视觉卡片压缩:投影片每页信息量上限、按依赖切页的动态规划与卡片字数」(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-151 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026151,
author = {Fan, Yafu},
title = {Content Fission Algebra: Compiling Hard-Core Monographs into Two-Host Podcasts, Study Slides and Quiz Banks},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}