Unsupervised Domain Adaptation with Adversarial Residual Transform Networks

TL;DR

This paper introduces ARTN, a novel adversarial domain adaptation method that enhances generalization and training stability by transforming source features into target feature space using residual connections and specialized regularization.

Contribution

The paper proposes ARTN, a new adversarial domain adaptation approach with residual connections and regularization, improving generalization and training stability over existing methods.

Findings

ARTN achieves comparable performance to state-of-the-art methods.

The model demonstrates improved training stability.

Experimental results on multiple datasets validate effectiveness.

Abstract

Domain adaptation is widely used in learning problems lacking labels. Recent studies show that deep adversarial domain adaptation models can make markable improvements in performance, which include symmetric and asymmetric architectures. However, the former has poor generalization ability whereas the latter is very hard to train. In this paper, we propose a novel adversarial domain adaptation method named Adversarial Residual Transform Networks (ARTNs) to improve the generalization ability, which directly transforms the source features into the space of target features. In this model, residual connections are used to share features and adversarial loss is reconstructed, thus making the model more generalized and easier to train. Moreover, a special regularization term is added to the loss function to alleviate a vanishing gradient problem, which enables its training process stable. A series of experiments based on Amazon review dataset, digits datasets and Office-31 image datasets are conducted to show that the proposed ARTN can be comparable with the methods of the state-of-the-art.