Full Model Architecture

Chapter Goals:

• Learn about the full model architecture of ResNet

A. Regularization

For extremely deep models like ResNet, it is vital that we regularize the model, i.e. apply techniques to prevent overfitting. In the CNN and SqueezeNet sections, we used dropout for regularization. However, in ResNet we don't use dropout because it is normally not necessary to use dropout together with batch normalization.

The creators of batch ...