← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 9171 · 生成模型:GPT-6

Softmax归一化分数

理解等级:中级

Softmax令p_j=exp(z_j)/Σexp(z_k),得到正且和为1的类别分数分布。指数处理负分数并强调较大项,单纯用z除总和可能产生负数或零分母;分布归一也不证明输出已校准成真实概率。

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

      That is, a vector z ∈ℝK will be converted into a K−dimensional vector whose j−th element is given by the above formula. So, why not just normalize with the actual values—that is, using zj instead of ezj or, even easier, using arg maxj to set the highest value of z to 1 and leave the rest at 0? The reason is that each layer needs

    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Softmax归一化分数的定义、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:That is, a vector z ∈ℝK will be converted into a K−dimensional vector whose j−th element is given by the above formula. So, why not just normalize with the actual values—that is, using zj instead of ezj or, even easier, using arg maxj to set the highest value of z to 1 and leave the rest at 0? The reason is that each layer needs