Interlacing Families and Sums of Independent Random Matrices

Interlacing Families and Sums of Independent Random Matrices

Probability Seminar
Sep 17, 2014, 03:10 PM - 04:00 PM | 1011 Evans Hall | Happening As Scheduled
Nikhil Srivastava, Microsoft Research and U.C. Berkeley
There are many situations in which one is interested in showing that some random matrix is likely to be small in spectral norm. In the special case when the matrix is a sum of independent bounded rank one random matrices, the situation is quite well-understood: the ‘Matrix Chernoff bound’ and related inequalities tell us that the norm is unlikely to be larger than a quantity which depends...