知识卡片
簇均值除以簇内样本数
更新第j簇中心应为Σi1(ai=j)xi除以Nj,其中Nj是该簇样本数。书中式30.6把分母写成总样本数N,会将小簇中心错误地缩向零。若Nj=0,均值未定义,需要保留、重置中心或其他明确空簇策略。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 表示学习
原书出处 · 1 处
- · 查看原书依据
where i′ enumerates all the datapoints for which ai′ = j. Recall that the point that minimizes the sum of squared distances from a dataset is the mean of the dataset. This yields our update in equation (30.6): just set each code to be the mean of all datapoints assigned to that code's cluster.
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第30章《表示学习》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c030b.xhtml)
- 结论依据:摘录中关于簇均值除以簇内样本数的定义、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:where i′ enumerates all the datapoints for which ai′ = j. Recall that the point that minimizes the sum of squared distances from a dataset is the mean of the dataset. This yields our update in equation (30.6): just set each code to be the mean of all datapoints assigned to that code's cluster.