← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6677 · 生成模型:GPT-6

降维目标依保留什么信息而定

理解等级:中级

把高维点投到较少维度,会舍弃一些变化;不同投影可保留不同结构。用于可视化、存储、检索或分类的好坏标准并不一致,保留总体变化最大的方向也可能丢失少数类关键线索,不能称降维必定提升泛化。

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

      Before we dive into LSA, let’s take a moment to understand what, conceptually, it does to our data. The idea behind LSA’s approach to topic modeling is dimensionality reduction. As its name suggests, dimensionality reduction is a process in which we find a lower-dimensional representation of data that retains as much information as possible.

    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:摘录中关于降维目标依保留什么信息而定的定义、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:Before we dive into LSA, let’s take a moment to understand what, conceptually, it does to our data. The idea behind LSA’s approach to topic modeling is dimensionality reduction. As its name suggests, dimensionality reduction is a process in which we find a lower-dimensional representation of data that retains as much information as possible.