Timezone: »

Why does Throwing Away Data Improve Worst-Group Error?
Kamalika Chaudhuri · Kartik Ahuja · Martin Arjovsky · David Lopez-Paz

Tue Jul 25 09:18 PM -- 09:26 PM (PDT) @ Meeting Room 316 A-C

When facing data with imbalanced classes or groups, practitioners follow an intriguing strategy to achieve best results. They throw away examples until the classes or groups are balanced in size, and then perform empirical risk minimization on the reduced training set. This opposes common wisdom in learning theory, where the expected error is supposed to decrease as the dataset grows in size. In this work, we leverage extreme value theory to address this apparent contradiction. Our results show that the tails of the data distribution play an important role in determining the worst-group-accuracy of linear classifiers. When learning on data with heavy tails, throwing away data restores the geometric symmetry of the resulting classifier, and therefore improves its worst-group generalization.

Author Information

Kamalika Chaudhuri (UCSD, Meta AI Research, and FAIR)
Kartik Ahuja (FAIR (Meta AI))
Martin Arjovsky (New York University)
David Lopez-Paz (Facebook AI Research)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors