Principled and Scalable Methods for Extracting multivariate dependencies in Big Data
Add to Google Calendar
Making sense of the many complex relationships and multivariate dependencies in "Big Data", formulating correct models, and developing inferential procedures is one of the major challenges facing data scientists. To this end high dimensional graphical models have been very useful and have found widespread applications. A popular approach for learning high dimensional graphical models is to use L1 regularization methods to induce sparsity in the inverse covariance estimator, leading to sparse partial covariance/correlation graphs. One major gap in the area is that none of the popular approaches proposed outside the Gaussian setting lead to well-defined estimators. To address this, we propose a new regression based graphical model selection method that is tractable, leads to well defines estimators, and also has good large sample properties and computational complexity. The methodology is illustrated on both real and simulated data. (Joint work with S. Oh and K. Khare)
Bala Rajaratnam is a faculty at Stanford University in the Department of Statistics and Environmental Earth System Science. He is affiliated with several other departments including the Institute for computational and mathematical engineering, the financial and risk modeling institute, the cardiovascular institute at the Stanford School of Medicine, and the Woods Institute for the Environment. His research interests include Graphical models, machine learning, data science, high dimensional inference, signal processing, spatio-temporal and environmental modeling and financial engineering. He is the recipient of several awards and recognitions including two federal CAREER awards, the NSF CAREER Award and the DARPA Young Faculty Award, and also the school wide excellence in teaching award.