Humans have the innate capabilities to learn and apply knowledge in problem solving and decision-making. At the same time we also have intrinsic abilities of transferring knowledge between related tasks. It is observed that if the new task is closely related to the previous learning, humans can quickly transfer this knowledge to perform the new task. For example, knowledge gained in learning how to ride a bicycle is adapted in learning how to drive a motorcycle. In artificial intelligence, it is desired to utilize this capability to make the algorithms adaptable. However, given some prior knowledge in a related task, traditional algorithms do not adapt to a new task and have to learn the new task from the beginning. Generally, they do not consider that the two tasks may be related and the knowledge gained in one may be used to learn the new task efficiently in lesser time.
Domain Adaptation attempts to mimic this human behavior by transferring the knowledge learned in one or more source tasks and uses it for learning the related target task. Researchers are developing machine learning algorithms are which are adaptable and utilize the domain knowledge and previously learnt models. Advances in domain adaptation literature has benefited biometric research significantly and approaches such as co-training, transfer learning, and other supervised/unsupervised domain adaptation have helped in developing improved recognition algorithms. For example, learning from high resolution source domain images and transferring the knowledge to learning low-resolution target domain information has helped in building cross-resolution face recognition algorithms. This tutorial will focus on the foundations, theory, recent advances on domain adaptation, and their applications in biometrics. Specifically, co-training, transfer learning, online (incremental/decremental) learning, covariate shift, domain adaptation in representation learning, shared representation learning, multimodal learning, and evolutionary computation based domain-adaptation algorithms will be discussed.
Richa Singh received the Ph.D. degree in Computer Science from West Virginia University, Morgantown, USA, in 2008. She is currently an Associate Professor with the IIIT Delhi, India and an Adjunct Associate Professor at West Virginia University, USA. Her areas of interest are biometrics, pattern recognition, and machine learning. She is a recipient of the Kusum and Mohandas Pai Faculty Research Fellowship at the IIIT Delhi, the FAST Award by Department of Science and Technology, India, and several best paper and best poster awards in international conferences. She has published over 200 research papers in journals, conferences and book chapters. She is also an Editorial Board Member of Information Fusion (Elsevier) and Associate Editor of IEEE Access and the EURASIP Journal on Image and Video Processing (Springer). She has also served as the Program Co-Chair of IEEE BTAS 2016 and General Co-Chair of ISBA 2017.
Mayank Vatsa received the Ph.D. degree in Computer Science from West Virginia University, Morgantown, USA, in 2008. He is currently an Associate Professor with the IIIT Delhi, India and an Adjunct Associate Professor and Visiting Professor at West Virginia University, USA. His areas of interest are biometrics, image processing, computer vision, and information fusion. He is a recipient of the AR Krishnaswamy Faculty Research Fellowship, the FAST Award by DST, India, and several best paper and best poster awards in international conferences. He has published more than 200 peer-reviewed papers in journals and conferences. He is also the Vice President (Publications) of IEEE Biometrics Council, an Associate Editor of the IEEE Access, and an Area Editor of Information Fusion (Elsevier). He has also served as the PC Co-Chair of ICB 2013, IJCB 2014, and ISBA 2017.