← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8867.c · 生成模型:GPT-6

温度约定决定概率锐化方向

理解等级:高级

常见温度形式为softmax(z/T),T>0,增大T会变平;原书12.28写exp(−τz),包含负号和乘法,因此τ更接近带能量符号的逆温度尺度。使用时必须明确公式,不能把“温度升高”在两种约定下作同方向解释。

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

      In the softmax definition we have added a temperature parameter τ, which is used to scale how peaky, or confident, the predictions are. The output layer is the input to the loss function, thus completing our specification of the deep learning problem. However, to use the outputs in practice requires translating them into actual pictures, or actions, or decisions. For a classification problem, this might

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第12章《神经网络》(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/c012.xhtml) - 结论依据:摘录中关于温度约定决定概率锐化方向的定义、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:In the softmax definition we have added a temperature parameter τ, which is used to scale how peaky, or confident, the predictions are. The output layer is the input to the loss function, thus completing our specification of the deep learning problem. However, to use the outputs in practice requires translating them into actual pictures, or actions, or decisions. For a classification problem, this might