知识卡片

普通读书笔记卡 · 6682.a · 生成模型:GPT-6

误差范数与均方根差一个维度因子

理解等级:中级

差向量的L2范数为各坐标差平方求和再开方,RMSE还要先除以维数;平方欧氏则不开方。L1、Jaccard、编辑距离和Mahalanobis各需不同对象或统计条件,不能只凭都叫距离便互换解释。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      Euclidean/Cartesian distance or root mean squared error (RMSE) —2-norm or L2 Squared Euclidean distance; sum of squares distance (SSD) —L2 norm squared Cosine or angular or projected distance —Normalized dot product Minkowski distance —p-norm or Lp Fractional distance, fractional norm —p-norm or Lp for 0 < p < 1 City block, Manhattan, or taxicab distance; sum of absolute distance (SAD) —1-norm or L1 Jaccard distance;

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第04章《从词数到语义分析》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-4.html) - 结论依据:摘录中关于误差范数与均方根差一个维度因子的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:Euclidean/Cartesian distance or root mean squared error (RMSE) —2-norm or L2 Squared Euclidean distance; sum of squares distance (SSD) —L2 norm squared Cosine or angular or projected distance —Normalized dot product Minkowski distance —p-norm or Lp Fractional distance, fractional norm —p-norm or Lp for 0 < p < 1 City block, Manhattan, or taxicab distance; sum of absolute distance (SAD) —1-norm or L1 Jaccard distance;