How to Build CNNs

Components of CNNs

• Convolution Layer
• Pooling Layer
• FC Layer
• Normalization Layer
• Dropout
• Activation Function

Normalization

Checkout Normalization: What, When, Why, and How?

Dropout

During Traning, in each forward pass, randomly set some neurons to zero; usually 0.5 (half of the neuron)

intuition: makes the nn model more robust

In test time, we usually use all the neuron, thus we need to scale it.