知识卡片
高维检索需结合稀疏性选方法
词表维数高会带来存储和几何度量问题,但并不意味着超过某个固定维数就必须用近似近邻。词项稀疏性可支持倒排索引与精确候选计算,稠密嵌入又有不同代价;应按数据规模、召回要求和索引实测选择。
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
原书出处 · 1 处
- · 查看原书依据
No matter how many dimensions you have, the linear algebra works the same. So you can use the vector math in this chapter to measure interesting things about your vectors. You will eventually run into the curse of dimensionality, but in chapters 4 and 10, you will use advanced algorithms to avoid that curse. Having too many dimensions is a curse because high-dimensional vectors will
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第03章《词频与逆文档频率向量》(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-3.html)
- 结论依据:摘录中关于高维检索需结合稀疏性选方法的定义、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:No matter how many dimensions you have, the linear algebra works the same. So you can use the vector math in this chapter to measure interesting things about your vectors. You will eventually run into the curse of dimensionality, but in chapters 4 and 10, you will use advanced algorithms to avoid that curse. Having too many dimensions is a curse because high-dimensional vectors will