关于「间隔复习排程:最佳间隔随保持期增长、莱特纳盒子的吸收链与每日复习上限」,《专著改编的信息保留代数》给出的定量结论是什么?

结论:保守排程的平均答对率最高,0.94,每张卡片只需 5.7856 次复习,每天 115.7 次;但它毕业时只证明了卡片能在 5 天后被回忆,按「每证明一天的复习次数」算是 1.1571,效率最低。加倍排程的平均答对率 0.90,每张卡 6.9351 次复习,每天 138.7 次,每证明一天 0.4334 次。激进排程答对率降到 0.82,每张卡 9.5008 次,每天 190.0 次,每证明一天 0.2969 次。鲁莽排程答对率只有 0.69,每张卡 17.1856 次复习,其中 12.1856 次是答错,每天 343.7 次。

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。

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: 0.2969, 0.4334, 0.69, 0.82, 0.90, 0.94

schedulepassreviewsmissesload/dayper proven day
timid 1,1,2,3,50.945.78560.7856115.71.1571
doubling 1,2,4,8,160.906.93511.9351138.70.4334
bold 2,4,8,16,320.829.50084.5008190.00.2969
reckless 4,8,16,32,640.6917.185612.1856343.70.2685
查看完整验算程序
# Chapter 5. p52: daily review load of a Leitner box schedule.
# A quiz card sits in box b = 1 to 5. In box b its memory stability is
# 2^(b - 1) days and it is reviewed after the box interval I_b; it is
# answered with probability 1 / (1 + I_b / (9 S_b)), the power law of
# p51. A pass moves it up one box, a miss sends it back to box 1, and a
# pass in box 5 retires it. The chain is absorbing; expected reviews and
# days per card follow from E_b = 1 + p_b E_(b+1) + (1 - p_b) E_1,
# solved exactly by back-substitution. With 20 new cards a day the
# steady daily load is 20 times the expected reviews (Little's law).
# A retired card has been recalled after I_5 days, its proven horizon.
# Schedules and stabilities illustrative.
import math

NEW_PER_DAY = 20
SCHEDULES = (("timid 1,1,2,3,5", (1, 1, 2, 3, 5)),
             ("doubling 1,2,4,8,16", (1, 2, 4, 8, 16)),
             ("bold 2,4,8,16,32", (2, 4, 8, 16, 32)),
             ("reckless 4,8,16,32,64", (4, 8, 16, 32, 64)))


def pass_prob(interval, box):
    return 1.0 / (1.0 + interval / (9.0 * 2.0 ** box))


def expected(costs, probs):
    # value_b = alpha_b + beta_b * value_1, built from box 5 down
    alpha, beta = 0.0, 0.0
    for c, p in zip(reversed(costs), reversed(probs)):
        alpha, beta = (math.fsum((c, p * alpha)),
                       math.fsum((p * beta, 1.0 - p)))
    return alpha / (1.0 - beta)


print("schedule               pass  reviews   misses  load/day  "
      "per proven day")
for name, iv in SCHEDULES:
    probs = [pass_prob(i, b) for b, i in enumerate(iv)]
    reviews = expected([1.0] * 5, probs)
    print("{:21s}  {:4.2f}  {:7.4f}  {:7.4f}  {:8.1f}  {:14.4f}".format(
        name, math.fsum(probs) / 5, reviews, reviews - 5.0,
        NEW_PER_DAY * reviews, reviews / iv[-1]))

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 5 章「间隔复习排程:最佳间隔随保持期增长、莱特纳盒子的吸收链与每日复习上限」(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}
}

本书其他章节结论