知识卡片

普通读书笔记卡 · 9257.a · 生成模型:GPT-6

输出归一不等于概率已校准

理解等级:中级

Softmax提供总和为1的类别分数表达,但预测0.9不自动意味着同类情况正确率90%。是否可当可靠概率仍需数据和校准证据;互斥多类与可同时多标签也不能套同一输出解释。

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

      Let us break down and understand each part of this equation. Notice that the input to this activation function is a bold y, which denotes a vector rather than a single number. It can have an arbitrary number of elements (e.g., y =[y1, y2, ... yn], where each yi is a single number). Now, the softmax function operates by exponentiating each element of the input

    参考来源

    - 位置:《Machine Learning Fundamentals: Concepts, Models, and Applications》第08章《深度学习模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Machine Learning Fundamentals: Concepts, Models, and Applications/_epub-src/OEBPS/xhtml/21_chapter08.xhtml) - 结论依据:摘录中关于输出归一不等于概率已校准的定义、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:Let us break down and understand each part of this equation. Notice that the input to this activation function is a bold y, which denotes a vector rather than a single number. It can have an arbitrary number of elements (e.g., y =[y1, y2, ... yn], where each yi is a single number). Now, the softmax function operates by exponentiating each element of the input