知识卡片

普通读书笔记卡 · 6676.h · 生成模型:GPT-6

精确率与召回率分母不同

理解等级:中级

精确率为TP/(TP+FP),描述报为正类的结果有多少正确;召回率为TP/(TP+FN),描述真正正类找到了多少。示例TP=125、FP=913、FN=201,分别约12.0%与38.3%;后续表格将这两项交换,不能照抄。

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

      First of all, out of 326 comments in the test set that were actually toxic, the model was able to identify only 125 correctly—that’s 38.3%. This measure (how many of the instances of the class we’re interested in the model was able to identify) is called recall, or sensitivity. On the other hand, out of 1,038 comments the model labeled as toxic, only 125 are

    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:摘录中关于精确率与召回率分母不同的定义、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:First of all, out of 326 comments in the test set that were actually toxic, the model was able to identify only 125 correctly—that’s 38.3%. This measure (how many of the instances of the class we’re interested in the model was able to identify) is called recall, or sensitivity. On the other hand, out of 1,038 comments the model labeled as toxic, only 125 are