Tutorial on Domain Adaptation


April 29, 2011


Hal Daumé III


University of Maryland


Almost anyone who has deployed machine learning systems in the real world has encountered the task of domain adaptation: We build our models from some fixed source domain, but we wish to deploy them across one or more different target domains. For example, large-scale speech recognition systems need to work well across arbitrary speech, regardless of background noise or accents. Text processing systems trained on news often need to be applied to blogs or forum posts. Gene finders are trained on a particular organism, but often we wish to identify the genes of another organism or even group of organisms. Face recognition systems might be trained under certain pose, lighting, and occlusion settings, but applied to arbitrary sets of pose, lighting, and occlusion. The purpose of this tutorial is to introduce participants to the problem of domain adaptation, the variety of forms it takes, the techniques that have been used to solve it, and our current understanding of when these techniques can and cannot work. We hope that our tutorial leads to new and interesting work on the open questions of domain adaptation.


Hal Daumé III

Hal Daumé III is an assistant professor in Computer Science at the University of Maryland, College Park. He holds joint appointments in UMIACS and Linguistics. His primary research interest is in understanding how to get human knowledge into a machine learning system in the most efficient way possible. He works primarily in the areas of language (computational linguistics and natural language processing) and machine learning (structured prediction, domain adaptation and Bayesian inference). He associates himself most with conferences like ACL, ICML, NIPS and EMNLP, and has over 30 conference papers (one best paper award in ECML/PKDD 2010) and 7 journal papers. He earned his PhD at the University of Southern California with a thesis on structured prediction for language (his advisor was Daniel Marcu). He spent the summer of 2003 working with Eric Brill in the machine learning and applied statistics group at Microsoft Research. Prior to that, he studied math (mostly logic) at Carnegie Mellon University. He still likes math and doesn’t like to use C (instead he uses O’Caml or Haskell). He doesn’t like shoes, but does like activities that are hard on your feet: skiing, badminton, Aikido and rock climbing.