Michael I. Jordan

Michael I. Jordan has played a seminal role in the development of statistical machine learning. His work, and the work of his former students and postdocs, has served to define this area, which bridges computer science and statistics. Michael played a leading role in the development of the variational approach to inference in graphical models, identifying techniques in statistical physics that could provide efficient approximation algorithms for large-scale Bayesian networks. Michael and his colleagues developed the Latent Dirichlet Allocation model for extracting topics from documents, which has proven very influential not only in text processing but in many other domains. He and his colleagues have also developed several important families of probability models, including the widely used hierarchical mixtures of experts, and the widely applicable hierarchical Dirichlet processes. These models have extraordinarily broad applications from robotics and computer vision, to text and natural language processing, to the detection of bugs in software, to statistical genomics and proteomics.

Michael's work is not only highly influential in computer science, with broad applicability within and beyond the field, but also plays a major role in the increasingly computational nature of research in statistics. Michael has also served as a leader in the machine learning community, organizing conferences and workshops, including serving as general chair of the NIPS conference. He is known for working broadly with neuroscientists, statisticians, and biologists, not only in joint research but as an educator and ambassador. Michael has trained an incredible cohort of students and postdocs, distinguished both in size and quality, who now hold faculty positions in over a dozen of the world's top computer science departments and positions in many leading companies.