← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6679 · 生成模型:GPT-6

奇异值分解需核对因子维数

理解等级:中级

对m×n的实矩阵W,完整SVD写作U(m×m)Σ(m×n)Vᵀ(n×n)。保留k个方向后写作Uₖ(m×k)Σₖ(k×k)Vₖᵀ(k×n)。维数明确后才能相乘;原书截断公式右因子的下标混乱,不应以图式标签代替核对。

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

      Let’s say we have our dataset, consisting of m n-dimensional points, represented by a matrix W. In its full version, this is what SVD of W would look like in math notation (assuming m > n):

    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:摘录中关于奇异值分解需核对因子维数的定义、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:Let’s say we have our dataset, consisting of m n-dimensional points, represented by a matrix W. In its full version, this is what SVD of W would look like in math notation (assuming m > n):