1 min read

Class imbalance: bug or feature?

Medical training, of necessity, over-represents unusual medical conditions. There are lots of unusual medical conditions, and trainees need to see at least a reasonable sampling of them. It is proverbial that new doctors tend to over-classify in favour of unusual medical conditions: in the 1940’s a medical professor said “When you hear hoofbeats behind you, don’t expect to see a zebra.”

In machine learning, there is a lot of concern about class imbalance. If