知识卡片

普通读书笔记卡 · 9086.a · 生成模型:GPT-6

混淆矩阵要说明归一化方向

理解等级:中级

按真类行归一化的C_ij表示真类i被预测为j的比例,各行总和为一。其对角线是各类召回率,不是预测该类的精确率;总体准确率需按类别样本量加权,不能在类别不平衡时直接平均对角元素。

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

      where Ci,j measures the percentage of times that true class i is classified as class j. In this toy example (table 50.2), we can see that 80 percent of the cat images are correctly classified, while 5 percent of cat images are classified as car images, and 15 percent are classified as pictures of dogs. 50.3.4 Shortcomings The task of image classification is plagued with

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第50章《目标识别》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c050b.xhtml) - 结论依据:摘录中关于混淆矩阵要说明归一化方向的定义、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:where Ci,j measures the percentage of times that true class i is classified as class j. In this toy example (table 50.2), we can see that 80 percent of the cat images are correctly classified, while 5 percent of cat images are classified as car images, and 15 percent are classified as pictures of dogs. 50.3.4 Shortcomings The task of image classification is plagued with