Juan K. Lin

Short Bio:

I received a Ph.D. in statistical physics at the University of Chicago, and was an
Applied Math Instructor in the Mathematics Department at the Massachusetts Institute of
Technology. Currently I am an Assistant Professor of Statistics at Rutgers University.

My research interests are in the fields of data mining, machine learning and information retrieval.
Recent projects focus on:

Network analysis of large sparse graphs based on traffic models. Applications in graph and relational analysis, and low rank approximations of Markov chains.
Parallel (coupled) implementations of the EM algorithm for parameter estimation in multinomial mixture models.
Model construction based on partial information. Approximate (combinatorial) inference using information propagation on a lattice-ordered set.

Select Papers:

Traffic Models for Community-based Ranking and Navigation, to appear, Lecture Notes in Computer Science: Workshop on Internet and Network Economics, Springer-Verlag.
Probabilistic Multi-State Split-Merge Algorithm for Coupling Parameter Estimates, Proceedings of the 2006 SIAM Conference on Data Mining (SDM'2006).
Lattice Theoretic Aspects of Inference and Parameter Estimation, to appear, Bayesian Inference and Maximum Entropy Methods in Science and Engineering, 2005.
Latent variable models for probabilistic graph clustering, Bayesian Inference and Maximum Entropy Methods in Science and Engineering, 24, pp. 187--194, 2004.
Approximate inference based on convex set sampling", in Bayesian and Maximum-Entropy Methods 23.
Latent variable models for link analysis of similarity data, in 35th Symposium on the Interface: Computing Science and Statistics.
Reduced-rank approximations of transition matrices, in Artificial Intelligence and Statistics, 2003, Key West, FL.
Curved gaussian models with application to the modeling of foreign exchange rates, in Proceedings of the Sixth International Conference on Computational Finance.