Stochastic Learning on Imbalanced Data: https://arxiv.org/abs/1705.00607v1
We study a mini-batch diversification scheme for stochastic gradient descent (SGD). While classical SGD relies on uniformly sampling data points to form a mini-batch, we propose a non-uniform sampling scheme based on the Determinantal Point Process (DPP). The DPP relies on a similarity measure between data points and gives low probabilities to mini-batches which contain redundant data, and higher probabilities to mini-batches with more diverse data. This simultaneously balances the data and leads to stochastic gradients with lower variance. We term this approach Balanced Mini-batch SGD (BM-SGD). We show that regular SGD and stratified sampling emerge as special cases. Furthermore, BM-SGD can be considered a generalization of stratified sampling to cases where no discrete features exist to bin the data into groups. We show experimentally that our method results more interpretable and diverse features in unsupervised setups, and in better classification accuracies in supervised setups.
We proposed a diversified mini-batch sampling scheme based on determinantal point processes. Our method, BMSGD, builds on a similarity matrix between the data points and suppresses the co-occurance of similar data points in the same mini-batch. This leads to a training procedure which generalizes better to more balanced test sets. We also derived sufficient conditions under which the method reduces the variance of the stochastic gradient, leading to faster learning. We showed that our approach generalizes both stratified sampling and pre-clustering. In the future, we will explore the possibility to further improve the efficiency of the algorithm with data reweighing and tackle imbalanced learning problems involving different modalities for supervised and multi-modal settings.
