知识卡片

普通读书笔记卡 · 8872.b · 生成模型:GPT-6

反向线性只针对固定工作点

理解等级:高级

在固定前向输入和参数后,局部雅可比是确定矩阵,反向传播对传入的上游梯度呈线性。改变前向值会改变这些矩阵,因此不能说参数到最终梯度的整个映射也是线性;若继续求高阶导数,还必须追踪雅可比对前向值的依赖。

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

      the backward network only consists of linear layers. This is true no matter what the forward network consists of (even if it is not a conventional neural network but some arbitrary computation graph). The reason why this happens is because backprop implements the chain rule, and the chain rule is always a product of Jacobian matrices. Since a Jacobian is a matrix, clearly it is

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第14章《反向传播》(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/c014.xhtml) - 结论依据:摘录中关于反向线性只针对固定工作点的定义、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:the backward network only consists of linear layers. This is true no matter what the forward network consists of (even if it is not a conventional neural network but some arbitrary computation graph). The reason why this happens is because backprop implements the chain rule, and the chain rule is always a product of Jacobian matrices. Since a Jacobian is a matrix, clearly it is