关于「东西方原典考辨:《孟子·尽心下》与高斯的最小二乘」,《可穿戴读数与广告日报的卡尔曼滤波》给出的定量结论是什么?

结论:第七章考辨《孟子·尽心下》的「尽信书,则不如无书」与高斯 1809 年的最小二乘。

数据口径:本书用局部水平与局部趋势的状态空间模型检验这一常识:一个读数等于真实水平加测量误差,当误差方差远大于真实日变化方差时,按原始读数做是或否的决定会对噪音过度反应;示意参数下原始读数每百天翻转 23.6 次,卡尔曼滤波估计只翻转 5.5 次。 书中的噪音方差、漂移、回报率与预算均为示意参数;全部计算由书中程序生成。

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

own nbiasw*RMSE w=1RMSE w=0RMSE w*
10.00.8001.0002.0000.894
11.00.8001.4142.0001.200
13.00.8003.1622.0002.561
40.00.5001.0001.0000.707
41.00.5001.4141.0000.866
43.00.5003.1621.0001.658
280.00.1251.0000.3780.354
281.00.1251.4140.3780.375
283.00.1253.1620.3780.515
查看完整验算程序
# Chapter 7. p72: "Believing the book entirely" written as a gain.
# A stated value (the "book": a device's stated accuracy, a platform's
# reported lift, a published average) has error with sd TAU, and may be
# biased by B. The reader has n own readings, each with error sd 2.0.
# Estimate = w * book + (1 - w) * mean(readings). w = 1 is believing
# the book entirely, w = 0 is having no book, and the Kalman weight is
# w* = (r / n) / (r / n + TAU^2) with TAU = 1.0 assumed. RMSE is exact:
# sqrt(w^2 (TAU^2 + B^2) + (1 - w)^2 r / n). Illustrative numbers.
import math

R, TAU = 4.0, 1.0


def rmse(w, n, bias):
    return math.sqrt(w * w * (TAU ** 2 + bias ** 2)
                     + (1 - w) ** 2 * R / n)


print("own n  bias  w*     RMSE w=1   RMSE w=0   RMSE w*")
for n in (1, 4, 28):
    ws = (R / n) / (R / n + TAU ** 2)
    for bias in (0.0, 1.0, 3.0):
        print("%5d  %4.1f  %.3f  %8.3f   %8.3f   %8.3f"
              % (n, bias, ws, rmse(1.0, n, bias), rmse(0.0, n, bias),
                 rmse(ws, n, bias)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《卡尔曼滤波状态空间同化、可穿戴读数观测噪音与广告归因真实信号提纯母典》第 7 章「东西方原典考辨:《孟子·尽心下》与高斯的最小二乘」(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-245 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026245,
  author    = {Fan, Yafu},
  title     = {Kalman Filtering and State-Space Assimilation: Separating Observation Noise from True Signals in Wearable Readings and Advertising Reports},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论