Advances achieved by deep learning-based domain adaptation have witnessed many real-world successful applications that include image classification, semantic segmentation, image generation, text classification, etc. In addition, various techniques like model training strategy and model architecture design have been proposed over the past years for the implementation of deep learning-based domain adaptation algorithms in the above applications. Generally, it is difficult to train or optimize deep learning-based domain adaptation methods whose essences are neural networks with many layers since the required number of training samples for obtaining satisfactory performance is rapidly increasing.v

However, by reducing divergences of distributions across domains, it is possible to enlarge the number of training samples. Additionally, by utilizing neural networks’ excellent capability of learning abstract representations, deep learning-based domain adaptation methods mine and transfer the common knowledge across tasks, which in turn can promote their respective performance. Recent studies have shown that the performance of deep learning-based domain adaptation methods has some dependence on model training strategies and architecture design. However, finding an optimal network architecture or parameters for deep learning-based domain adaptation methods may not be easy.

This Special Issue intends to gather the latest high-quality original research and review articles discussing deep learning-based domain adaptation methods and applications. Meanwhile, we welcome articles that focus on the network architecture design of deep learning-based domain adaptation models, the strategies of model training, their explanation and visualization of performance evaluation metrics, etc.

Potential topics include but are not limited to the following:

• Deep learning-based domain adaptation methods for image classification, semantic segmentation, style-transfer, image translation, text classification, image caption generation, neural machine translation, etc
• Deep learning-based domain adaption with different levels of supervised knowledge, including supervised domain adaptation, semi-supervised domain adaption, and unsupervised domain adaptation
• Theoretical analyses of deep learning-based domain adaptation algorithms, including how to reduce the gap between distributions of the domains, the learning of domain-invariant feature learning or mapping of domains, etc
• Comprehensive performance comparison between the conventional domain adaptation algorithms and deep learning-based domain adaptation algorithms
• The exploration of model training strategies for deep learning-based domain adaptation methods
• Network architecture searching or the model design for deep learning-based domain-adaptation methods
• The combination of deep learning-based domain adaptation and generative model including generative adversarial networks, variational auto-encoding or adversarial auto-encoder, etc
• Application of deep learning-based domain adaptation methods in other potential fields such as financial engineering, medical image analysis, health-care electronic records, face verification, time series analysis, etc