An introduction to domain adaptation and transfer learning

Abstract

In machine learning, if the training data is an unbiased sample of an underlying distribution, then the learned classification function will make accurate predictions for new samples. However, if the training data is not an unbiased sample, then there will be differences between how the training data is distributed and how the test data is distributed. Standard classifiers cannot cope with changes in data distributions between training and test phases, and will not perform well. Domain adaptation and transfer learning are sub-fields within machine learning that are concerned with accounting for these types of changes. Here, we present an introduction to these fields, guided by the question: when and how can a classifier generalize from a source to a target domain? We will start with a brief introduction into risk minimization, and how transfer learning and domain adaptation expand upon this framework. Following that, we discuss three special cases of data set shift, namely prior, covariate and concept shift. For more complex domain shifts, there are a wide variety of approaches. These are categorized into: importance-weighting, subspace mapping, domain-invariant spaces, feature augmentation, minimax estimators and robust algorithms. A number of points will arise, which we will discuss in the last section. We conclude with the remark that many open questions will have to be addressed before transfer learners and domain-adaptive classifiers become practical.

arXiv commentTechnical Report. 41 pages, 5 figures

References in corpus (14)

• SMOTE: Synthetic Minority Over-sampling Technique
• Rényi Divergence and Kullback-Leibler Divergence
• Domain Generalization via Invariant Feature Representation
• Marginalized Denoising Autoencoders for Domain Adaptation
• Domain Adaptation: Learning Bounds and Algorithms
• Game theory, maximum entropy, minimum discrepancy and robust Bayesian decision theory
• Data Mining and Machine Learning in Astronomy
• Efficient Learning of Domain-invariant Image Representations
• Sample-to-Sample Correspondence for Unsupervised Domain Adaptation
• Analysis of Kernel Mean Matching under Covariate Shift
• A New PAC-Bayesian Perspective on Domain Adaptation
• Learning an MR acquisition-invariant representation using Siamese neural networks
• MR Acquisition-Invariant Representation Learning
• Target contrastive pessimistic risk for robust domain adaptation

Cited by in corpus (41)

• A Comprehensive Survey of Deep Transfer Learning for Anomaly Detection in Industrial Time Series: Methods, Applications, and Directions
• New high-quality strong lens candidates with deep learning in the Kilo Degree Survey
• Lifelong Teacher-Student Network Learning
• Understanding and Improving Information Transfer in Multi-Task Learning
• Training Variational Networks with Multi-Domain Simulations: Speed-of-Sound Image Reconstruction
• Supervised Domain Adaptation: A Graph Embedding Perspective and a Rectified Experimental Protocol
• Domain Adaptation and Autoencoder Based Unsupervised Speech Enhancement
• Cross-Domain Few-Shot Learning by Representation Fusion
• Shift-Robust GNNs: Overcoming the Limitations of Localized Graph Training Data
• Bridging the gap: Towards an Expanded Toolkit for AI-driven Decision-Making in the Public Sector
• Bridging Physics-based and Data-driven modeling for Learning Dynamical Systems
• DC-Check: A Data-Centric AI checklist to guide the development of reliable machine learning systems
• Self-supervised Assisted Active Learning for Skin Lesion Segmentation
• Transfer learning of phase transitions in percolation and directed percolation
• Developing Personalized Models of Blood Pressure Estimation from Wearable Sensors Data Using Minimally-trained Domain Adversarial Neural Networks
• A Two-Sample Conditional Distribution Test Using Conformal Prediction and Weighted Rank Sum
• MetaCI: Meta-Learning for Causal Inference in a Heterogeneous Population
• EllSeg-Gen, towards Domain Generalization for head-mounted eyetracking
• Fair Classification via Domain Adaptation: A Dual Adversarial Learning Approach
• A Primer on Domain Adaptation
• Coping with Label Shift via Distributionally Robust Optimisation
• SurvITE: Learning Heterogeneous Treatment Effects from Time-to-Event Data
• Score-based Generative Neural Networks for Large-Scale Optimal Transport
• Energy-constrained Self-training for Unsupervised Domain Adaptation
• Deepfake Videos in the Wild: Analysis and Detection
• Adversarial Sample Enhanced Domain Adaptation: A Case Study on Predictive Modeling with Electronic Health Records
• Unsupervised Domain Adaptation across FMCW Radar Configurations Using Margin Disparity Discrepancy
• Assessing Machine Learning Approaches to Address IoT Sensor Drift
• Domain Generalization under Conditional and Label Shifts via Variational Bayesian Inference
• Transfer learning of individualized treatment rules from experimental to real-world data
• MetaXT: Meta Cross-Task Transfer between Disparate Label Spaces
• Optimal Transfer Learning Model for Binary Classification of Funduscopic Images through Simple Heuristics
• Adversarial Training for Multi-domain Speaker Recognition
• Causal Learning for Socially Responsible AI
• A general approach to bridge the reality-gap
• Does the dataset meet your expectations? Explaining sample representation in image data
• Optimal Ensemble Construction for Multi-Study Prediction with Applications to COVID-19 Excess Mortality Estimation
• Making Table Understanding Work in Practice
• Learning to Transfer: A Foliated Theory
• Recursively Conditional Gaussian for Ordinal Unsupervised Domain Adaptation
• Generating Annotated Training Data for 6D Object Pose Estimation in Operational Environments with Minimal User Interaction