I am not clear about the best way to use regularization in image classification. One option I assume is to change the wd (weight decay) parameter value - are there other parameters that can be tuned - for example batch norm or dropout if applicable. I searched the docs and the forums but couldn’t find any examples on how to add these - so example code on how to do this would be great.
Also some context - I am trying this on an image dataset that is highly imbalanced, so I have implemented oversampling, but it is still overfitting, I was thinking increasing regularization may help. Any other ideas are welcome too.