深度学习-How to design Conv Network

• How to design Conv Network:

• The first thing to understand is what is the sensory eye

• first conv a pink pixel represents a 3x3 perceived region

• Second conv a pink pixel represents a 5x5 perceived region

• 

• 

• Three convolutional layers are superimposed, and the final sensory eye range is 7x7

• 

• Fewer parameters, more nonlinearities, and fewer computations

• 

• 1x1 is actually the core point of the residual network  3.25 is not true  6C^2

• 

• The main idea of designing a network

• 

•