Dear all,
On Tuesday, November 13, Prof. Alvaro R. De Pierro will give a talk
at the mathematics colloquium at the University of Haifa. The talk
starts at 16:10. Snacks and coffee are served at 16:00. The title and
abstract of his talk are ((in the attached pdf)))
(( or see them below))
Best regards,
Frol Zapolsky
Speaker: Alvaro R. De Pierro,
         Applied Mathematics Department, University of Campinas, Brazil
Most inverse problems could be modeled by optimization of functions that are de-
composed as a sum of two parts : one related to the consistency of the equations
that should be in principle satisfied and the other to prior information, not contai-
ned in those equations. Recently, the explosion of articles on Compressed Sensing
(CS) increased the importance of considering nondifferentiable priors and specific
iterative methods for the solution of the associated optimization problems. Most
of these methods are based on classical general methods for nondierentiable opti-
mization. In this talk we emphasize on a dierent family of methods that consider
the fact that we are dealing with functions that represent measures of information
and the advantages of that approach. We analyze the algorithms from two different
and complementary points of view : the majorizing functions approach [1] and its
corresponding geometrical interpretation in an appropriate space as a sequence of
projections in a given measure [2], [3]. Our reference inverse problem will be Emission
Computed Tomography.
[1] A.R. De Pierro, A modified expectation maximization algorithm for penali-
zed likelihood estimation in emission tomography" IEEE Transactions on Medical
Imaging, V14, 1, 132-137, 1995.
[2] Eggermont PPB, LaRiccia VN , On EM- like algorithms for minimum distance
estimation. \ University of Delaware 1998.
[3] Byrne C. and Eggermont P., EM Algorithms - in Handbook for Mathematical
Methods in Imaging, Springer Verlag, ed O. Scherzer 273-337, 2011.
Technion Math Net-2 (TECHMATH2)
Editor: Michael Cwikel   <> 
To see today's activities and other future and past activities go to
Announcement from: Frol Zapolsky   <>