Imbalanced Data 
==================


Subsampling
---------------
Sample from the more frequent class less frequently

Weighting 
------------
Weight the loss of the less frequent class more