Deep Adversarial Domain Adaptation Based on Multi-layer Joint Kernelized Distance

TL;DR

This paper introduces a deep adversarial domain adaptation method that leverages multi-layer joint kernelized distance to select and utilize target data for improved adaptation performance across different data distributions.

Contribution

It proposes a novel multi-layer joint kernelized distance metric combined with a class-balanced pseudo-labeling strategy within an adversarial framework for domain adaptation.

Findings

Outperforms several state-of-the-art methods in domain adaptation tasks.

Effectively selects target data likely to be correctly classified.

Enhances adaptation by using target data as labeled through pseudo labels.

Abstract

Domain adaptation refers to the learning scenario that a model learned from the source data is applied on the target data which have the same categories but different distribution. While it has been widely applied, the distribution discrepancy between source data and target data can substantially affect the adaptation performance. The problem has been recently addressed by employing adversarial learning and distinctive adaptation performance has been reported. In this paper, a deep adversarial domain adaptation model based on a multi-layer joint kernelized distance metric is proposed. By utilizing the abstract features extracted from deep networks, the multi-layer joint kernelized distance (MJKD) between the $j$th target data predicted as the $m$th category and all the source data of the $m'$th category is computed. Base on MJKD, a class-balanced selection strategy is utilized in each category to select target data that are most likely to be classified correctly and treat them as labeled data using their pseudo labels. Then an adversarial architecture is used to draw the newly generated labeled training data and the remaining target data close to each other. In this way, the target data itself provide valuable information to enhance the domain adaptation. An analysis of the proposed method is also given and the experimental results demonstrate that the proposed method can achieve a better performance than a number of state-of-the-art methods.