← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6677.d · 生成模型:GPT-6

文档坐标与变换基各司其职

理解等级:中级

若W以文档为行、词为列,低维文档坐标为WVₖ,理论上对应UₖΣₖ;新文档行向量q也乘Vₖ。components_通常存Vₖᵀ,即主题×词项,变换时需其转置;它本身不是文档×主题的输出。

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

      You can think of V as a “transformer” tool used to map your data from the “old” space (its representation in matrix W’s “world”) to the new, lower-dimensional one. Imagine we added a few more points to our 3D horse point cloud and now want to understand where those new points would be in our 2D representation, without needing to recalculate the transformation for all

    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:摘录中关于文档坐标与变换基各司其职的定义、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:You can think of V as a “transformer” tool used to map your data from the “old” space (its representation in matrix W’s “world”) to the new, lower-dimensional one. Imagine we added a few more points to our 3D horse point cloud and now want to understand where those new points would be in our 2D representation, without needing to recalculate the transformation for all