知识卡片
句向量归一需处理零范数
每个非零句向量除以其L2范数后,向量间点积即余弦相似度。全未知词或互相抵消可能产生零向量,需要跳过或另定义处理;归一后统一长度并不意味着所有特征独立、语义或尺度问题都解决。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 用词嵌入推理
原书出处 · 1 处
- · 查看原书依据
Now that you have sentence vectors and noun phrases, you should normalize61 the sentence vectors so that all your vectors have a length (or 2-norm) of 1. The 2-norm is computed the same way you compute the length of the diagonal across a right triangle: you add up the square of the length of the dimensions, and then you take the square root of the
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:摘录中关于句向量归一需处理零范数的定义、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:Now that you have sentence vectors and noun phrases, you should normalize61 the sentence vectors so that all your vectors have a length (or 2-norm) of 1. The 2-norm is computed the same way you compute the length of the diagonal across a right triangle: you add up the square of the length of the dimensions, and then you take the square root of the