A model trained on the raw distribution may optimize aggregate loss while performing poorly on rare classes. Resampling, class-weighted losses, threshold selection, and class-specific metrics can address the imbalance.
Class imbalance occurs when some target classes have many more examples than others.
A model trained on the raw distribution may optimize aggregate loss while performing poorly on rare classes. Resampling, class-weighted losses, threshold selection, and class-specific metrics can address the imbalance.