Dresden 2014 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

DY: Fachverband Dynamik und Statistische Physik

DY 22: Statistical Physics (general)

DY 22.3: Talk

Wednesday, April 2, 2014, 15:30–15:45, ZEU 160

Dynamics in Stochastic Optimization: Combining Stochastic Tunneling and Energy Landscape Paving — •Kay Hamacher — Depts. of Physics, Computer Science, and Biology, TU Darmstadt

Heuristic optimization schemes such as simulated annealing, genetic algorithms, or extremal optimization play a most prominent role in global optimization. The performance of these algorithms and their respective sampling behavior during the search process are themselves interesting problems - in particular from the viewpoint of dynamical systems theory and statistical mechanics.

Here, we show that a combination of two approaches * namely Energy Landscape Paving (ELP) and Stochastic Tunneling (STUN) * can overcome known problems of other Metropolis-sampling-based procedures. We show on grounds of non-equilibrium statistical mechanics and empirical evidence on the synergistic advantages of this combined approach and discuss simulations for a complex optimization problem.

Reference: [1] K. Hamacher. A New Hybrid Metaheuristic - Combining Stochastic Tunneling and Energy Landscape Paving, 8th International Workshop on Hybrid Metaheuristics (HM2013), Lecture Notes in Computer Science (LNCS 7919), pp. 107-117, 2013.