A Closer Look at Smoothness in Domain Adversarial Training

TL;DR

This paper investigates how smoothness in the optimization landscape affects domain adversarial training, revealing that smoothing task loss improves target domain performance while smoothing adversarial loss does not, leading to the proposed SDAT method.

Contribution

The paper provides a theoretical analysis of smoothness effects in domain adversarial training and introduces SDAT, a new method that enhances domain adaptation performance.

Findings

Smooth minima in task loss stabilize adversarial training.

Smoothing adversarial loss leads to sub-optimal generalization.

SDAT improves performance in classification and object detection.

Abstract

Domain adversarial training has been ubiquitous for achieving invariant representations and is used widely for various domain adaptation tasks. In recent times, methods converging to smooth optima have shown improved generalization for supervised learning tasks like classification. In this work, we analyze the effect of smoothness enhancing formulations on domain adversarial training, the objective of which is a combination of task loss (eg. classification, regression, etc.) and adversarial terms. We find that converging to a smooth minima with respect to (w.r.t.) task loss stabilizes the adversarial training leading to better performance on target domain. In contrast to task loss, our analysis shows that converging to smooth minima w.r.t. adversarial loss leads to sub-optimal generalization on the target domain. Based on the analysis, we introduce the Smooth Domain Adversarial Training (SDAT) procedure, which effectively enhances the performance of existing domain adversarial methods for both classification and object detection tasks. Our analysis also provides insight into the extensive usage of SGD over Adam in the community for domain adversarial training.