原书出处
Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)
Figure 14.11: Pytorch-like pseudocode for a linear layer with forward and backward. 14.6.2 Backpropagation for a Pointwise Nonlinearity Pointwise nonlinearities have very simple backward functions. Let a (parameterless) scalar nonlinearity be h : ℝ → ℝ with derivative function h′ : ℝ → ℝ. Define a pointwise layer using h as f (xin) = [h(xin[0]), …, h(xin[N – 1])]T. Then we have (14.30) The diag
原书定位详情
来源:a19c7933-4eda-4176-aebd-30fe7e1ad292
正文位置:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c014.xhtml
EPUB SHA-256:60a7eebc43f5d9a675d776fed3736fbdaa791e55877023f2f8cba016a9b384f8