Feature maps
Why not Linear
- 4 Layers: [784, 256, 256, 256, 10]
335k or 1.3MB
em...
- 486 PC + AT&T DSP32C
Batch X
Gradient Cache
etc.
Receptive Field
Fully connnected
Partial connected
Locally connected
Rethink Linear layer
Fully VS Lovally
Weight sharing
Why call Convolution？
2D Convolution
\[
y(t) = x(t)*h(t) = \int_{-\infty}^{\infty}x(\tau)h(t-\tau)d\tau
\]
Convolution in Computer Vision
CNN on feature maps
原文链接：http://www.cnblogs.com/nickchen121/p/10923353.html
经验首页