知识卡片

普通读书笔记卡 · 8939.c · 生成模型:GPT-6

输出门控制记忆读出

理解等级:中级

更新细胞状态后,输出门oₜ决定哪些内容作为隐藏态传出:hₜ=oₜ⊙tanh(cₜ)。细胞里保留的信息可以暂时不完全暴露在当前hₜ中。对h作限制不代表c内容已经删除,读取、保留和写入分别由不同计算路径控制。

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

      Finally, given the updated cell state, we read from it and use the values to determine the hidden state to output: (25.11) (25.12) The basic idea is it should be easy, by default, not to forget. The cell state controls what is remembered. It gets modulated, but ft can learn to be all ones so that information propagates forward through an identity connection from the

    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:摘录中关于输出门控制记忆读出的定义、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:Finally, given the updated cell state, we read from it and use the values to determine the hidden state to output: (25.11) (25.12) The basic idea is it should be easy, by default, not to forget. The cell state controls what is remembered. It gets modulated, but ft can learn to be all ones so that information propagates forward through an identity connection from the