Difference between revisions of "Rodolphe Sepulchre, June 2013"
Line 18:  Line 18:  
=== Abstract ===  === Abstract ===  
+  
+  <center>  
+  '''The geometry of (thin) SVD revisited for largescale computations'''  
+  
+  Rodolphe Sepulchre<br>  
+  University of Liege, Belgium  
+  </center>  
+  
+  The talk will introduce a riemannian framework for largescale computations over the  
+  set of lowrank matrices. The foundation is geometric and the motivation  
+  is algorithmic, with a bias towards efficient computations in largescale problems.  
+  We will explore how classical matrix factorizations connect the riemannian geometry of the set of  
+  fixedrank matrices to two wellstudied manifolds: the Grassmann manifold of linear subspaces and the cone  
+  of positive definite matrices. The theory will be illustrated on various applications, including  
+  lowrank Kalman filtering, linear regression with lowrank priors, matrix completion, and the choice of a suitable 
Revision as of 10:25, 28 May 2013
Rodolphe Sepulchre will visit Caltech on 3 June 2013 (Mon).
Agenda
9:30a  Richard Murray, 109 Steele Lab 
9:45a  Meet with Richard's NCS group, 110 Steele

11:45  Seminar setup 
12:00p  Seminar, 213 ANB 
1:00p  Lunch with Venkat, CMS faculty 
2:15p  Open 
3:00p  Open 
3:45p  Open 
4:30p  Open 
5:15p  Done 
Abstract
The geometry of (thin) SVD revisited for largescale computations
Rodolphe Sepulchre
University of Liege, Belgium
The talk will introduce a riemannian framework for largescale computations over the set of lowrank matrices. The foundation is geometric and the motivation is algorithmic, with a bias towards efficient computations in largescale problems. We will explore how classical matrix factorizations connect the riemannian geometry of the set of fixedrank matrices to two wellstudied manifolds: the Grassmann manifold of linear subspaces and the cone of positive definite matrices. The theory will be illustrated on various applications, including lowrank Kalman filtering, linear regression with lowrank priors, matrix completion, and the choice of a suitable