← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6706.a · 生成模型:GPT-6

亲和矩阵的代价随句数平方增长

理解等级:中级

N个D维归一向量组成S,S乘Sᵀ得到对称N×N相似矩阵,点积对称并非一般矩阵乘法可交换。存储约O(N²),计算约O(N²D);先建全稠密矩阵再筛边,仍需承受此前峰值成本。

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

      With the sentence vectors normalized, you can get the similarity between all those vectors and each other. Computing the pairwise similarity between all the possible pairs of objects in a list of object creates a square matrix, called a similarity matrix or affinity matrix, as shown in the following listing.63 If you use the dot product of each vector with all the others, you are

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第06章《用词嵌入推理》(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-6.html) - 结论依据:摘录中关于亲和矩阵的代价随句数平方增长的定义、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:With the sentence vectors normalized, you can get the similarity between all those vectors and each other. Computing the pairwise similarity between all the possible pairs of objects in a list of object creates a square matrix, called a similarity matrix or affinity matrix, as shown in the following listing.63 If you use the dot product of each vector with all the others, you are