Prof. Emo Welzl and Prof. Bernd Gärtner
|Mittagsseminar Talk Information|
Date and Time: Tuesday, December 04, 2012, 12:15 pm
Duration: 30 minutes
Location: CAB G51
Speaker: Sebastian Stich
A continuous function f is to be minimized. Similar as in the discrete setting, people often use Randomized Search Heuristics for this task. While they often give good results in practice, most of them lack a thorough theoretical convergence analysis.
In this talk, we present and analyze Variable Metric Random Pursuit. This is an iterative algorithm where in every step a new approximation is calculated by searching along a randomly chosen direction. However, typically not every direction yields the same progress. The algorithm tries to "learn" successful search directions and samples them more often.
Automatic MiSe System Software Version 1.4803M | admin login