Prof. Emo Welzl and Prof. Bernd Gärtner
|Mittagsseminar Talk Information|
Date and Time: Friday, July 11, 2014, 12:15 pm
Duration: 30 minutes
Location: OAT S15/S16/S17
Speaker: Christian Mueller (New York University)
I review a number of L1-regularized statistical estimators, LASSO, sparse inverse covariance selection (SPICS), and TREX, their corresponding (non-)convex optimization problems, and their numerical solution in practice. While LASSO and SPICS are well-known tools in high-dimensional statistics that require an independent selection of a tuning parameter, the TREX is a novel method that is tuning-free. I show the applicability of these estimators in a diverse range of systems biology applications. The applications are concerned (i) with sparse interaction graph recovery from microbial population abundance data, (ii) high-accuracy prediction of detrimental point mutations in proteins from sequence and structure information, and (iii) high-dimensional classification and regression from genomics and proteome data.
This joint work with members of the Bonneau lab, Courant Institute and Center for Genomics and Systems Biology, NYU and Johannes Lederer, Jacob Bien, and Irina Gayanova, Department of Statistics, Cornell University.
Automatic MiSe System Software Version 1.4803M | admin login