关于「细节分档:固定专注时长里的多选背包」,《瓶颈定焦与价值工时比排程》给出的定量结论是什么?

结论:全部精做的结果最能说明问题:八件只做完三件,总价值 101,是最优安排的 0.783。全部粗做八件都做完了,总价值只有 58,是最优的 0.450。全部中做做完七件,总价值 114,是最优的 0.884,已经不坏,但仍差一截。最优安排的总价值是 129。

数据口径:史密斯 1956 年证明,单人面对一张清单,使加权延迟代价最小的次序是按价值除以耗时排序,而不是按价值排序;七件示意任务上,按重要性排比穷举最优多付 0.311。 图形学的视锥剔除与细节层次说明,固定帧预算下不该每件都精做:八件示意任务全部精做只做完三件,总价值只有最优安排的 0.783。

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。 与结论吻合的打印值: 0.450, 0.783, 0.884, 101, 114, 129

rule value tasks done
all coarse58 8 of 8 (0.450 of optimum)
all medium114 7 of 8 (0.884 of optimum)
all fine101 3 of 8 (0.783 of optimum)
LOD optimum129
chosen levels:
K1 checkout copy fine100 min
K2 consult reply fine70 min
P1 chapter edit medium80 min
P2 cover brief medium40 min
V1 video script coarse30 min
V2 thumbnail coarse10 min
S1 restock sheet coarse20 min
S2 ledger notes coarse10 min
minutes used 360 of 360
查看完整验算程序
# Chapter 5. p51: level of detail as a multiple-choice knapsack.
# A focus budget of 360 minutes (breaks are budgeted separately, see
# chapter 6). Eight tasks, each done at one level: skip, coarse, medium
# or fine, with (minutes, value). Values are concave in minutes; the two
# checkout-node tasks (K1, K2) lose most of their value when done
# coarsely. Exact dynamic programme over 10-minute units versus three
# uniform rules: all coarse, all medium, all fine (uniform rules take
# tasks in w/p order of that level until the budget runs out).
# All minutes and values are illustrative.
TASKS = [  # name, [(min, val) coarse, medium, fine]
    ("K1 checkout copy", [(30, 8), (60, 30), (100, 44)]),
    ("K2 consult reply", [(20, 6), (40, 22), (70, 30)]),
    ("P1 chapter edit", [(40, 14), (80, 22), (160, 27)]),
    ("P2 cover brief", [(20, 6), (40, 9), (90, 11)]),
    ("V1 video script", [(30, 10), (60, 15), (120, 18)]),
    ("V2 thumbnail", [(10, 4), (20, 6), (50, 7)]),
    ("S1 restock sheet", [(20, 7), (40, 10), (80, 11)]),
    ("S2 ledger notes", [(10, 3), (30, 5), (60, 6)]),
]
BUDGET = 360
LEVELS = ["skip", "coarse", "medium", "fine"]


def dp():
    n, cap = len(TASKS), BUDGET // 10
    best = [[-1.0] * (cap + 1) for _ in range(n + 1)]
    pick = [[0] * (cap + 1) for _ in range(n + 1)]
    best[0] = [0.0] * (cap + 1)
    for i, (_, opts) in enumerate(TASKS):
        choices = [(0, 0)] + opts
        for b in range(cap + 1):
            for lv, (m, v) in enumerate(choices):
                u = m // 10
                if u <= b and best[i][b - u] + v > best[i + 1][b]:
                    best[i + 1][b] = best[i][b - u] + v
                    pick[i + 1][b] = lv
    lv, b = [], cap
    for i in range(n, 0, -1):
        lv.append(pick[i][b])
        b -= ([(0, 0)] + TASKS[i - 1][1])[pick[i][b]][0] // 10
    return best[n][cap], lv[::-1]


def uniform(level):
    def ratio(i):
        m, v = TASKS[i][1][level]
        return -v / m
    order = sorted(range(len(TASKS)), key=ratio)
    left, val, used = BUDGET, 0, 0
    for i in order:
        m, v = TASKS[i][1][level]
        if m <= left:
            left -= m
            val += v
            used += 1
    return val, used


opt, lv = dp()
print("rule          value  tasks done")
for k, name in ((0, "all coarse"), (1, "all medium"), (2, "all fine")):
    v, u = uniform(k)
    print("%-12s  %5d  %d of 8   (%.3f of optimum)" % (name, v, u, v / opt))
print("LOD optimum   %5d" % opt)
print("chosen levels:")
mins = 0
for (name, opts), l in zip(TASKS, lv):
    m = ([(0, 0)] + opts)[l][0]
    mins += m
    print("  %-18s %-6s %3d min" % (name, LEVELS[l], m))
print("minutes used %d of %d" % (mins, BUDGET))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《阶段瓶颈动态定焦、价值工时比优先级排序与认知视锥渲染加速母典》第 5 章「细节分档:固定专注时长里的多选背包」(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-376 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026376,
  author    = {Fan, Yafu},
  title     = {Stage-Bottleneck Triage, Weighted Shortest-Processing-Time Scheduling and Level-of-Detail Execution Velocity},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论