[NeurIPS 2022] Source code for our paper "Escaping Saddle Points for Effective Generalization on Class-Imbalanced Data" - View it on GitHub