← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8938.a · 生成模型:GPT-6

长雅可比链导致梯度失衡

理解等级:高级

跨T步的导数包含多个状态雅可比矩阵连乘。ReLU简单RNN的局部雅可比为RₜW,其中Rₜ由当前激活门控决定;连乘可能在某些方向放大、另一些方向衰减。不能仅凭W各元素看起来大或小判断,矩阵谱、奇异值、非正规性和门控路径都重要。

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

      Notice that the we can get very large computation graphs when we run an RNN for many timesteps. The gradient computation involves a series of matrix multiplies that is O(T) in length. For example, to compute our gradient in the example above involves the following product:

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第25章《循环神经网络》(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/c025.xhtml) - 结论依据:摘录中关于长雅可比链导致梯度失衡的定义、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:Notice that the we can get very large computation graphs when we run an RNN for many timesteps. The gradient computation involves a series of matrix multiplies that is O(T) in length. For example, to compute our gradient in the example above involves the following product: