Linear Digressions

Machine learning on imbalanced classes: surprisingly tricky. Many (most?) algorithms tend to just assign the majority class label to all the data and call it a day. SMOTE is an algorithm for manufacturing new minority class examples for yourself, to help your algorithm better identify them in the wild. Relevant links: https://www.jair.org/media/953/live-953-2037-jair.pdf