知识卡片

普通读书笔记卡 · 9173.c · 生成模型:GPT-6

池化用局部摘要减少空间细节

理解等级:中级

最大池化取窗口最大激活,平均池化取平均值,窗口与步长决定重叠及降采样程度。平均池化是线性操作、最大池化非线性;局部摘要会丢失精细位置,不是只保留无损有用信息。

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

      A pooling operation applies a window to select the maximum (in which case it is referred to as MaxPooling) or the average, among many other possible nonlinear functions, from a window of a given size. Figure 10.11 shows the result of a MaxPooling layer with pool size of 3 × 3 and stride lengths of 1 × 1 and 3 × 3. Usually, the stride

    参考来源

    - 位置:《Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)》第10章《人工神经网络》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)/_epub-src/OEBPS/xhtml/chapter_10.xhtml) - 结论依据:摘录中关于池化用局部摘要减少空间细节的定义、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 pooling operation applies a window to select the maximum (in which case it is referred to as MaxPooling) or the average, among many other possible nonlinear functions, from a window of a given size. Figure 10.11 shows the result of a MaxPooling layer with pool size of 3 × 3 and stride lengths of 1 × 1 and 3 × 3. Usually, the stride