知识卡片
二元交叉熵包含正反两类
标签y∈{0,1}、正类预测p时,BCE=−y log p−(1−y)log(1−p)。一项对真实标签保留,另一项为零;多标签任务可有多个独立二元目标,不必把它们硬塞进互斥Softmax。
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
- · 人工神经网络
原书出处 · 1 处
- · 查看原书依据
In the case where there are only two classes, the binary cross-entropy is calculated as follows: As there are only two classes (either true or false), directly follows from . The more general case for N > 2 is known as categorical cross-entropy. When using one-hot encoding, only class c has probability 1 (yc = 1), reducing the cross-entropy to with c the true class
参考来源
- 位置:《Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)》第10章《人工神经网络》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)/_epub-src/OEBPS/xhtml/chapter_10.xhtml)
- 结论依据:摘录中关于二元交叉熵包含正反两类的定义、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:In the case where there are only two classes, the binary cross-entropy is calculated as follows: As there are only two classes (either true or false), directly follows from . The more general case for N > 2 is known as categorical cross-entropy. When using one-hot encoding, only class c has probability 1 (yc = 1), reducing the cross-entropy to with c the true class