← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6715 · 生成模型:GPT-6

最大池化保留强响应而丢位置

理解等级:中级

全局最大池化为每个输出通道保留全序列最大值,可形成不依赖序列长度的固定宽度表示。它强调是否存在强局部模式,但会丢掉具体位置、次数和其他响应;多处弱证据也未必由单个最大值表示。

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

      Step 3 is to decide whether the text contains a good match somewhere within it. For this, your CNN converts the sequence of values output by convolution into a single value. The result is a single value representing the likelihood that the kernel’s pattern was somewhere in the text. Most CNNs are designed to take the maximum value of this sequence of numbers as a

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第07章《卷积网络识别文本模式》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-7.html) - 结论依据:摘录中关于最大池化保留强响应而丢位置的定义、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:Step 3 is to decide whether the text contains a good match somewhere within it. For this, your CNN converts the sequence of values output by convolution into a single value. The result is a single value representing the likelihood that the kernel’s pattern was somewhere in the text. Most CNNs are designed to take the maximum value of this sequence of numbers as a