Tid: 2 november 2009 kl 15.15-16.00
Plats : Seminarierummet 3733, Institutionen för matematik, KTH, Lindstedts väg 25, plan 7. Karta!
Föredragshållare: Yuwei Zhao
Titel: Numerical algorithms for a class of obstacle problems (Examensarbete – Master thesis)
This paper concerns the nonsmooth Newton method and its applications in solving obstacle problems. A detailed description of the nonsmooth Newton method and its corresponding convergence theorem is presented. We show that Howard's algorithm generates the same sequence as the nonsmooth Newton method does in solving obstacle problems. We also consider a nonsmooth Newton method based on the so-called Fischer-Burmeister (FB) function. The penalty approximations are used to solve obstacle problems with the nonsmooth Newton method. Obstacle problems that arise in financial mathematics such as pricing American options, pricing American warrants and portfolio selection problems, are taken as examples to examine the algorithms and numerical tests are performed.
|Sidansvarig: Filip Lindskog