Efficient Augmentation for Imbalanced Deep Learning

TL;DR

This paper introduces an efficient three-phase CNN training framework with a novel data augmentation method, EOS, to improve classification accuracy on imbalanced datasets by reducing the generalization gap for minority classes.

Contribution

It proposes a new training framework and EOS augmentation technique that enhance minority class recognition while being computationally efficient.

Findings

Improved accuracy over existing resampling methods.

EOS reduces the generalization gap for minority classes.

Framework is more efficient than SMOTE and GAN-based oversampling.

Abstract

Deep learning models tend to memorize training data, which hurts their ability to generalize to under-represented classes. We empirically study a convolutional neural network's internal representation of imbalanced image data and measure the generalization gap between a model's feature embeddings in the training and test sets, showing that the gap is wider for minority classes. This insight enables us to design an efficient three-phase CNN training framework for imbalanced data. The framework involves training the network end-to-end on imbalanced data to learn accurate feature embeddings, performing data augmentation in the learned embedded space to balance the train distribution, and fine-tuning the classifier head on the embedded balanced training data. We propose Expansive Over-Sampling (EOS) as a data augmentation technique to utilize in the training framework. EOS forms synthetic training instances as convex combinations between the minority class samples and their nearest enemies in the embedded space to reduce the generalization gap. The proposed framework improves the accuracy over leading cost-sensitive and resampling methods commonly used in imbalanced learning. Moreover, it is more computationally efficient than standard data pre-processing methods, such as SMOTE and GAN-based oversampling, as it requires fewer parameters and less training time.