知识卡片
主成分按剩余方差依次选方向
PCA对中心化数据选择相互正交的方向,使保留的方差尽可能大,并按贡献排序。方差依赖特征尺度,保留比例可帮助确定维数,但不自动等于保留同样比例的语义或分类信息;是否先标准化需按特征单位与目标决定。
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
- · 从词数到语义分析
原书出处 · 1 处
- · 查看原书依据
To summarize, to reduce the number of dimensions describing your data without losing information, you need to find a representation that maximizes the variance along each of its new axes, while reducing the dependence between the dimensions and getting rid of those with high covariance. This is exactly what PCA does, by finding a set of dimensions along which the variance is maximized. These dimensions
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:摘录中关于主成分按剩余方差依次选方向的定义、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:To summarize, to reduce the number of dimensions describing your data without losing information, you need to find a representation that maximizes the variance along each of its new axes, while reducing the dependence between the dimensions and getting rid of those with high covariance. This is exactly what PCA does, by finding a set of dimensions along which the variance is maximized. These dimensions