| HOME | RESUME |
BIOGRAPHY | COURSE
INFORMATION | RESEARCH |
LINKS Biography My undergraduate training was in Mathematics and Logic at the University of Minnesota. I spent 1 year as a graduate student in the Math Department at UC Berkeley before being persuaded by Michele Loeve (who was in the statistics department at that time) to join the Department of Statistics. My father, Milton Sobel, also played a role in this transition. Initially, my interests consisted , almost exclusively in Probability Theory and Asymptotics. I later switched to working in the area of Decision Theory under the tutelage of Erich Lehmann. My thesis concerned a result involving the (surprising) admissibility of a class of estimators for Exponential Family parameters. I got married to a fellow Berkeley alum named Marilyn Carter. My first professional appointment was at the University of Georgia where I continued working in the area of Decision Theory, eventually getting interested in the related area's of Relevant Estimation, Applications of Statistics to Genetics (with my colleague Jonathon Arnold), and Shrinkage Estimation. In 1988 I left Georgia for Temple University in Philadelphia. Gradually, my statistical interests shifted, during this transition, to the area of Bayesian statistical inference. At this time I became interested in questions related to Bayesian modeling of sports and intellectual ability. During this period my wife and I had two wonderful children (named Ben and Elizabeth). After this, my interests shifted to research studying and incorporating Markov Chain Monte Carlo sampling, random forests, and recursive partitioning and related clustering/classification techniques. During this period I also did some joint work with Bud Mishra (Bioinformatics Unit, NYU). This lead to my work on a series of papers and grants with Rolf Lakaemper and others. We employed EM, MCMC, and many techniques to do data-mining, clustering, and scaling. It also resulted in joint work on many papers with Vasilis Megalooikonomou (CIS, Temple University) on topics related to shape recognition in MRI (magnetic resonance imaging), classification, etc.. We employed Random Forests, Recursive Partitioning, Mathematical/Statistical Morphology, data depth, MCMC, and Graphical Models to compare fMRI 'objects' . Finally I got involved in a great deal of work involving shape correspondence with Rolf Lakaemper. I've given invited talks at IEEE data mining (2003), IEEE KDD (2006) (on Kullback Leibler divergence), and CVPR (2008) (shape correspondence). I've recently gotten involved with Iyad Obeid doing work in the area of neural firing (see also my CV) At present my interests consist in Neural Firing, Shape Correspondence, sequential markov chain methodology, and multiscale random fields. I'm currently working with Rolf Lakaemper and a variety of other members of the Computer Information group (Temple University) on a variety of problems. a) Shape Correspondence b) Neural Firing c) sequential markov chain monte carlo d) Improved EM estimation using Kullback Leibler Divergence
|