Return of Frustratingly Easy Domain Adaptation

TL;DR

This paper introduces CORAL, a simple yet effective unsupervised domain adaptation method that aligns second-order statistics of source and target data, significantly improving performance without needing target labels.

Contribution

The paper presents CORAL, a straightforward, computationally efficient approach for unsupervised domain adaptation that outperforms more complex methods in benchmark tests.

Findings

CORAL is easy to implement, requiring only four lines of code.

CORAL achieves competitive results on standard benchmarks.

It effectively reduces domain shift without target labels.

Abstract

Unlike human learning, machine learning often fails to handle changes between training (source) and test (target) input distributions. Such domain shifts, common in practical scenarios, severely damage the performance of conventional machine learning methods. Supervised domain adaptation methods have been proposed for the case when the target data have labels, including some that perform very well despite being "frustratingly easy" to implement. However, in practice, the target domain is often unlabeled, requiring unsupervised adaptation. We propose a simple, effective, and efficient method for unsupervised domain adaptation called CORrelation ALignment (CORAL). CORAL minimizes domain shift by aligning the second-order statistics of source and target distributions, without requiring any target labels. Even though it is extraordinarily simple--it can be implemented in four lines of Matlab code--CORAL performs remarkably well in extensive evaluations on standard benchmark datasets.