Research Biography of Michael Jordan

Michael Jordan

Over the last twenty-five years, Michael Jordan has pioneered some of the most highly cited and most influential computational models of learning, inference and control in both biological and machine systems.  His work has brought advanced statistical thinking to the study of both machine learning and human learning, and helped to make the fields of statistics and psychology far more computationally sophisticated.  In this interdisciplinarity, he is the consummate cognitive scientist.

His early work on neural networks made multiple seminal contributions. These include the eponymous “Jordan networks”, the introduction of forward models for motor learning (joint work with Rumelhart), and some of the first quantitative evidence for optimal probabilistic inference, evidence integration and adaptation in sensorimotor systems (in three papers with his students and postdocs published in Science and Nature).  Of particular significance was his development of the  “Mixture of Experts” model, in joint work with Jacobs, Hinton and Nowlan, which demonstrated to the field of machine learning, as well as computationally sophisticated modelers of learning in cognitive science and cognitive neuroscience, the power of explicit probabilistic formulations and advanced methods for model fitting. This work pointed the way for thousands of researchers to follow in what has become a tremendously fertile multidisciplinary interaction between statistics, models of human learning, and algorithms and architectures for machine learning.

After moving from MIT to Berkeley in 1998, Jordan continued to focus on foundational issues in inference, learning and computation.  Over the subsequent years he has become a leading figure in the inferential fields of graphical models, Bayesian nonparametrics and machine learning.  An example is his work in these areas is his work with his student David Blei on Latent Dirichlet Allocation (LDA), or “topic models”, which has been among the most cited works in both computer science and statistics since its publication in 2003. This work gives a simple, tractable and surprisingly expressive probabilistic model for capturing the overall semantic content of sentences or documents.  It has had tremendous impact on computational models of semantic content in psycholinguistics and natural language processing, computational models of memory, and in both on the human and machine vision communities, for capturing the semantics of visual scenes.  More generally, Jordan’s work on Bayesian nonparametrics and graphical models has provided the theoretical underpinnings for models of categorization, word learning, perceptual learning, phonemic category acquisition and word segmentation and causal learning.  Most recently, together with Percy Liang and Dan Klein, Jordan has brought together advanced ideas from Bayesian inference with linguistically sophisticated representations to build computational models of semantics. In adapting and extending classic representations from formal semantics and discourse theory, embedded inside a state-of-the-art inferential framework, this work is cognitive science at its best.

Michael Jordan was one of Rumelhart’s students and a close personal friend, and he has carried on Rumelhart’s legacy of mentorship by himself mentoring a tremendous number of students and postdocs who have come to dominate and transform multiple fields. His research has epitomized the notion that the most important, deepest questions are not the exclusive province of any one discipline, and he has drawn broadly and voraciously on any and all approaches that could inform the answers to these questions, ranging from biology, physics, and mathematics to statistics and computer science.

Michael Jordan is Pehong Chen Distinguished Professor of Statistics and of Electrical Engineering and Computer Science at the University of California at Berkeley. After receiving his BS in Psychology from Louisiana State University and MS in Mathematics and Statistics at Arizona State University, he went on to pursue his PhD in Cognitive Science at the University of California, San Diego, under the supervision of David Rumelhart. After completing his degree in 1985, and a two-year postdoc at the University of Massachusetts at Amherst, Jordan was appointed assistant professor in the Department of Brain and Cognitive Sciences at MIT in 1988. In 1998, he left MIT for Berkeley where he joined the Computer Science and Statistics Departments.  He has received many honors, including his election as a member of both the National Academy of Sciences and the National Academy of Engineering, as well as being named fellow of numerous professional organizations, including the Cognitive Science Society.

Selected Publications

Jordan, M. I. & Rumelhart, D. E. (1992).  Forward models: Supervised learning with a distal teacher.  Cognitive Science, 16, 307-354.

Jordan, M. I. & Jacobs, R. A. (1994).  Hierarchical mixtures of experts and the EM algorithm.  Neural Computation, 6, 181-214.

Wolpert, D., Ghahramani, Z., & Jordan, M. I. (1995). An internal model for sensorimotor integration.  Science, 269, 1880-1880.

Saul, L., Jaakkola, T., & Jordan, M. I. (1996).  Mean field theory for sigmoid belief networks.  Journal of Artificial Intelligence Research, 4, 61-76.

Jordan, M. I., Ghahramani, Z., Jaakkola, T., & Saul, L. (1999).An introduction to variational methods for graphical models. Machine Learning, 37, 183-233.

Todorov, E. & Jordan, M. I. (2002). Optimal feedback control as a theory of motor coordination. Nature Neuroscience, 5, 1226-1235.

Blei, D., Ng, A., & Jordan, M. I. (2003).  Latent Dirichlet allocation.  Journal of Machine Learning Research, 3, 993-1022.

Jordan, M. I. (2004).  Graphical models.  Statistical Science, 19, 140-155.

Teh, Y. W., Jordan, M. I., Beal, M. J., & Blei, D. M. (2006). Hierarchical Dirichlet processes.  Journal of the American Statistical Association, 101, 1566-1581.

Wainwright, M. & Jordan, M. I. (2008).  Graphical models, exponential families and variational inference.  Foundations and Trends in Machine Learning, 1, 1-305.

Blei, D., Griffiths, T., & Jordan, M. I. (2010).  The nested Chinese restaurant process and Bayesian inference of topic hierarchies.  Journal of the ACM, 57, 1-30.

Liang, P., Jordan, M. I., & Klein, D. (2013).  Learning dependency-based compositional semantics.  Computational Linguistics, 39, 389-446.

Bouchard-Côté, A. & Jordan, M. I. (2013).  Evolutionary inference via the Poisson indel process.  Proceedings of the National Academy of Sciences, 110, 1160-1166.

Broderick, T., Jordan, M. I., & Pitman, J. (2013).  Cluster and feature modeling from combinatorial stochastic processes.  Statistical Science, 28, 289-312.

Jordan, M. I. (2013).  On statistics, computation and scalability. Bernoulli, 19, 1378-1390.