← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8936.a · 生成模型:GPT-6

时间步表示顺序计算

理解等级:初级

RNN的时间步是序列索引,不一定是物理时钟。它可以逐帧处理视频、逐像素扫描图像,或重复执行内部计算模块。设计递归前应明确每步接收什么、更新什么状态,以及状态何时初始化与重置;否则不同样本的历史可能混入同一状态。

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 idea of a time in an RNN does not necessarily mean time as it is measured on a clock. Time just refers to sequential computation; the timestep t is an index into the sequence. The sequential computation could progress over a spatial dimension (processing an image pixel by pixel) or over the temporal dimension (processing a video frame by frame), or even over more

    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:摘录中关于时间步表示顺序计算的定义、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 idea of a time in an RNN does not necessarily mean time as it is measured on a clock. Time just refers to sequential computation; the timestep t is an index into the sequence. The sequential computation could progress over a spatial dimension (processing an image pixel by pixel) or over the temporal dimension (processing a video frame by frame), or even over more