← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6676.l · 生成模型:GPT-6

线性与二次判别假设不同

理解等级:中级

线性判别分析对各类采用共享协方差,形成线性决策边界;二次判别允许各类不同协方差,边界可弯曲。更灵活同时增加估计需求,尤其小样本高维时未必更好;应按验证结果而非曲线更复杂选择。

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

      LDA will serve you well in many circumstances; however, it still has some assumptions that will cause the classifier to underperform when these assumptions are not fulfilled. For example, LDA assumes that the feature covariance matrices for all of your classes are the same. That’s a pretty strong assumption! As a result, LDA can only learn linear boundaries between classes.

    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:摘录中关于线性与二次判别假设不同的定义、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:LDA will serve you well in many circumstances; however, it still has some assumptions that will cause the classifier to underperform when these assumptions are not fulfilled. For example, LDA assumes that the feature covariance matrices for all of your classes are the same. That’s a pretty strong assumption! As a result, LDA can only learn linear boundaries between classes.