知识卡片
可训练滤波器与固定滤波器需区分
手工核固定定义要检测的局部模式,学得核则由目标损失引导寻找有用响应。两者都需合适输入和验证,训练不保证找到全部正确语法;若核未注册或被冻结,整条流水线有梯度也不代表核正在学习。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
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
Just a decade or so ago, before CNNs, you would have had to handcraft your kernels to match whatever patterns you could dream up. But with CNNs, you don’t have to program the kernels at all, except to decide how wide the kernels are—how many letters or words you think will capture the patterns you need. Your CNN optimizer will fill in the weights within
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:摘录中关于可训练滤波器与固定滤波器需区分的定义、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:Just a decade or so ago, before CNNs, you would have had to handcraft your kernels to match whatever patterns you could dream up. But with CNNs, you don’t have to program the kernels at all, except to decide how wide the kernels are—how many letters or words you think will capture the patterns you need. Your CNN optimizer will fill in the weights within