知识卡片
递归分开状态更新与读出
一般循环层写为hₜ=f(hₜ₋₁,xₜ),yₜ=g(hₜ)。f将新输入与已有记忆结合,g从当前状态产生任务输出,二者可选择不同函数。状态不必直接等于最终预测,分开二者有助于区分模型记住的信息和当前任务需要读出的结果。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 循环神经网络
原书出处 · 1 处
- · 查看原书依据
A recurrent layer is defined by the following equations: (25.1) (25.2) This is a layer that includes a state variable, h. The operation of the layer depends on its state. The layer also updates its state every time it is called, therefore the state is a kind of memory. The idea of a time in an RNN does not necessarily mean time as it is
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:摘录中关于递归分开状态更新与读出的定义、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:A recurrent layer is defined by the following equations: (25.1) (25.2) This is a layer that includes a state variable, h. The operation of the layer depends on its state. The layer also updates its state every time it is called, therefore the state is a kind of memory. The idea of a time in an RNN does not necessarily mean time as it is