IMB-NAS: Neural Architecture Search for Imbalanced Datasets

TL;DR

This paper introduces IMB-NAS, a neural architecture search method tailored for imbalanced datasets, focusing on efficient adaptation of architectures from balanced to long-tailed data to improve classifier performance.

Contribution

It proposes a novel approach to adapt NAS architectures for imbalanced datasets by retraining only the classification head, reducing computational costs.

Findings

Retraining the classification head with reweighted loss improves performance.

Freezing the backbone NAS super-network is effective for adaptation.

Extensive experiments validate the approach across multiple datasets.

Abstract

Class imbalance is a ubiquitous phenomenon occurring in real world data distributions. To overcome its detrimental effect on training accurate classifiers, existing work follows three major directions: class re-balancing, information transfer, and representation learning. In this paper, we propose a new and complementary direction for improving performance on long tailed datasets - optimizing the backbone architecture through neural architecture search (NAS). We find that an architecture's accuracy obtained on a balanced dataset is not indicative of good performance on imbalanced ones. This poses the need for a full NAS run on long tailed datasets which can quickly become prohibitively compute intensive. To alleviate this compute burden, we aim to efficiently adapt a NAS super-network from a balanced source dataset to an imbalanced target one. Among several adaptation strategies, we find that the most effective one is to retrain the linear classification head with reweighted loss, while freezing the backbone NAS super-network trained on a balanced source dataset. We perform extensive experiments on multiple datasets and provide concrete insights to optimize architectures for long tailed datasets.