知识卡片
多层感知机交替仿射与非线性
MLP用仿射层混合输入,用逐点非线性改变表示,再重复这一过程并映射到输出。隐藏单元可以区分激活前z和激活后h=g(z)。不同资料对层数计数不同,比较架构时应列出实际变换,不能只比较“几层”的名字。
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
- · 神经网络
原书出处 · 1 处
- · 查看原书依据
In general, MLPs can be constructed with any number of layers following this pattern: linear layer, activation function, linear layer, activation function, and so on. The activation function g could be the threshold function like in equation (12.2), but more generally it can be any pointwise nonlinearity, that is, and is any nonlinear function that maps ℝ → ℝ. Beyond MLPs, this kind of sequence
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第12章《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/c012.xhtml)
- 结论依据:摘录中关于多层感知机交替仿射与非线性的定义、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:In general, MLPs can be constructed with any number of layers following this pattern: linear layer, activation function, linear layer, activation function, and so on. The activation function g could be the threshold function like in equation (12.2), but more generally it can be any pointwise nonlinearity, that is, and is any nonlinear function that maps ℝ → ℝ. Beyond MLPs, this kind of sequence