← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6713.b · 生成模型:GPT-6

卷积响应是局部加权和

理解等级:中级

卷积核在序列上逐步滑动,对当前窗口各位置与通道乘权求和,形成新的响应序列。核可手工设计,也可由任务损失反馈学习;响应只是特征,不能直接称某种语义存在的概率。

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

      If you used a stencil this way, sliding it across the text to find the maximum match between your pattern and a piece of text, you’d be doing convolution with a stencil! When talking about deep learning and CNNs, the stencil is called a kernel or filter. In CNNs, a kernel is an array of floating-point numbers, rather than a cardboard cutout. The kernel is

    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:摘录中关于卷积响应是局部加权和的定义、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:If you used a stencil this way, sliding it across the text to find the maximum match between your pattern and a piece of text, you’d be doing convolution with a stencil! When talking about deep learning and CNNs, the stencil is called a kernel or filter. In CNNs, a kernel is an array of floating-point numbers, rather than a cardboard cutout. The kernel is