[ont.events] Solution of Discrete Regularization Problems by Means of Truncated SVD

vrsyrotiuk@water.waterloo.edu (Violet Syrotiuk) (04/03/89)

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

SCIENTIFIC COMPUTATION SEMINAR

                    - Thursday, April 6, 1989

Professor  Per  Christian Hansen*, of the University of
California  at Los Angeles, will speak on ``Solution of
Discrete  Regularization Problems by Means of Truncated
SVD''.

TIME:      1:30 PM

ROOM:      DC 1304

ABSTRACT

We  show  how  the  truncated SVD (TSVD) can be used to
compute regularized least-squares solutions to discrete
ill-posed  problems.  We also compare the TSVD solution
with   the  solution  computed  by  means  of  Tichonov
regularization,  and  we  give  perturbation bounds for
these  solutions.   An important result is that neither
of  the  methods  yields a reasonable solution unless a
discrete  Picard  condition  is satisfied.  In the last
part  of  the  talk,  we  try  to  give  an overview of
efficient   numerical  algorithms  for  computing  TSVD
solutions.
__________
*  On  leave  from  UNI.C, Technical University of
Denmark, DK-2800 Lyngby
-- 
Violet R. Syrotiuk     |                                   vrsyrotiuk@water.uucp
Computer Science Dept. |                                watmath!water!vrsyrotiuk
University of Waterloo |                           vrsyrotiuk@water.uwaterloo.ca
Waterloo, ON  N2L 3G1  |                 vrsyrotiuk@water.waterloo.edu (or .cdn)