A Review of Domain Adaptation without Target Labels

By Wouter M. Kouw, Marco Loog. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019

606 citations recorded by the source corpus.

Domain adaptation has become a prominent problem setting in machine learning and related fields. This review asks the question: How can a classifier learn from a source domain and generalize to a target domain? We present a categorization of approaches, divided into, what we refer to as, sample-based, feature-based, and inference-based methods. Sample-based methods focus on weighting individual observations during training based on their importance to the target domain. Feature-based methods revolve around on mapping, projecting, and representing features such that a source classifier performs well on the target domain and inference-based methods incorporate adaptation into the parameter estimation procedure, for instance through constraints on the optimization procedure. Additionally, we review a number of conditions that allow for formulating bounds on the cross-domain generalization error. Our categorization highlights recurring ideas and raises questions important to further research.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/3ef03001691a875ba5d11bdd2886bc962d94033e

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Federated learning for medical image analysis: A survey (2023, background): “Domain adaptation (Guan and Liu, 2022; Wilson and Cook, 2020; Kouw and Loog, 2019) is a sound machine learning technique that aims to reduce domain shift among different datasets and enhance the generalizability of a learning model.”
Deep Unsupervised Domain Adaptation: A Review of Recent Advances and Perspectives (2022, background): “In addition, some reviews do not touch deeply on domain mapping (Csurka, 2017; Kouw, 2018; Kouw and Loog, 2019), normalization statistic-based (Csurka, 2017; Kouw and Loog, 2019; Zhao et al.”
Domain Adaptation for Medical Image Analysis: A Survey (2021, neutral): “In terms of the number of source domains, the existing DA methods can be divided into single-source DA and multi-source DA approaches [12], [14], [15].”
Heterogeneous Graph Attention Network for Unsupervised Multiple-Target Domain Adaptation (2020, background): “Existing DA methods [3] can be divided into two main genres, semisupervised domain adaptation (SDA) and unsupervised domain adaptation (UDA) [4].”
Transfer learning: a friendly introduction (2022, background): “For instance, the Object-oriented paradigm algorithmically detects, recognizes, and describes the unseen images [48].”
Two standard-free approaches to correct for external influences on near-infrared spectra to make models widely applicable (2020, background): “The challenge of sensor calibration and model updating is not limited to the NIR spectroscopy domain as it is frequently encountered in other fields such as computer vision (Kouw and Loog, 2019).”
KinGDOM: Knowledge-Guided DOMain Adaptation for Sentiment Analysis (2020, positive): “Our work deals with domain-adversarial approaches (Kouw and Loog, 2019), where we extend DANN Ganin et al. (2016).”
A Comprehensive Survey on Test-Time Adaptation Under Distribution Shifts (2023, neutral): “Our survey is related to other surveys on domain adaptation [6], [18], [19], and prior adaptation [20].”
A Review of Single-Source Deep Unsupervised Visual Domain Adaptation (2020, background): “discussion devoted to recent deep learning-based methods; [27] covers almost all categories of domain adaptation methods briefly but comparisons are scant; [37, 38] focus on shallow methods and only a few deep methods are reviewed; [39, 40] analyze multi-source domain adaptation respectively”
Measuring Domain Shift for Deep Learning in Histopathology (2020, background): “Domain shift, deﬁned as differences between the source and target domain data, can be caused by, for example, differences due to sampling bias, differences in image content or view angle, or differences in image characteristics such as brightness, noise or color [4].”
Domain Adaptation by Joint Distribution Invariant Projections (2020, background): “Although matching the marginal distributions in this way has been justified theoretically [33], as mentioned in [4], [34], [35], this does not imply that the labeling distributions are matched as well.”
Maximum Density Divergence for Domain Adaptation (2020, background): “A typical domain adaptation setting consists of a welllabeled source domain and an unlabeled target domain [2], [9], [10].”