Delving into Deep Imbalanced Regression

TL;DR

This paper introduces Deep Imbalanced Regression (DIR), a new framework for handling continuous, imbalanced data in real-world tasks, proposing distribution smoothing techniques and providing benchmarks across multiple domains.

Contribution

The paper defines DIR for continuous targets, proposes distribution smoothing for labels and features, and provides large-scale benchmarks for practical imbalanced regression tasks.

Findings

Distribution smoothing improves regression performance on imbalanced data.

Benchmark datasets demonstrate the effectiveness of proposed methods.

Extensive experiments confirm the superiority of the strategies over existing approaches.

Abstract

Real-world data often exhibit imbalanced distributions, where certain target values have significantly fewer observations. Existing techniques for dealing with imbalanced data focus on targets with categorical indices, i.e., different classes. However, many tasks involve continuous targets, where hard boundaries between classes do not exist. We define Deep Imbalanced Regression (DIR) as learning from such imbalanced data with continuous targets, dealing with potential missing data for certain target values, and generalizing to the entire target range. Motivated by the intrinsic difference between categorical and continuous label space, we propose distribution smoothing for both labels and features, which explicitly acknowledges the effects of nearby targets, and calibrates both label and learned feature distributions. We curate and benchmark large-scale DIR datasets from common real-world tasks in computer vision, natural language processing, and healthcare domains. Extensive experiments verify the superior performance of our strategies. Our work fills the gap in benchmarks and techniques for practical imbalanced regression problems. Code and data are available at https://github.com/YyzHarry/imbalanced-regression.