February 07, 2017
4:30 PM
CORE, b-135
![](http://cdn.uclouvain.be/styles/full_content/groups/cms-editors-core/events/seminars/cover-operations-research-seminars.jpg?itok=YMWm2_te)
Derivative-free trust-region algorithms for composite nonsmooth optimization
Geovani NUNES GRAPIGLIA, University of Parana, Brazil
In this talk, we consider derivative-free trust-region algorithms to the problem of minimizing a composite function $\Phi(x)=f(x)+h(c(x))$, where $f$ and $c$ are smooth (and possibly nonconvex) and $h$ is convex but may be nonsmooth.This class of problems include L1 approximation problems and also the minimization of exact penalty functions corresponding to equality constrained optimization problems. Under certain conditions, global convergence and a function- evaluation complexity bound are proved. Numerical results are presented and the combination of these derivative-free algorithms and global solvers is illustrated in the context of parameter calibration of biochemical models.
Categories Events: