Hi all,
often times I see people saying that if you are underfitting/overfitting, that you should increase/decrease your models's complexity. However, what does this mean? Or rather, what specifically are we doing when we increase/decrease our model complexity? Any example prototxt files? Do we simply remove or add a layer? How would we do that while making sure the values (num_output, kernel size, stride, etc) are correct?