知识卡片

普通读书笔记卡 · 6677.f · 生成模型:GPT-6

降维后向量不会自动保持单位范数

理解等级:中级

TruncatedSVD输出连续的文档主题坐标,可能有负值,维数也低于原词表。它不自动将每行L2归一化;原TF–IDF已归一仍不保证投影后范数相同,后续若要余弦评分,需要重新核对。

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

      What you have just produced using the fit-transform method is your document vectors in the new representation. Instead of representing your comments with 19,169 frequency counts, you represented it with just 16. This matrix is also called a document-topic matrix. By looking at the columns, you can see how much every topic is “expressed” in every comment.

    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:摘录中关于降维后向量不会自动保持单位范数的定义、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:What you have just produced using the fit-transform method is your document vectors in the new representation. Instead of representing your comments with 19,169 frequency counts, you represented it with just 16. This matrix is also called a document-topic matrix. By looking at the columns, you can see how much every topic is “expressed” in every comment.