Evolutionary and Gradient-Based Algorithms for Lennard-Jones Cluster Optimization
S. Müller, N. N. Schraudolph, and P. Koumoutsakos. Evolutionary and Gradient-Based Algorithms for Lennard-Jones Cluster Optimization. In Genetic and Evolutionary Computation Conference Workshop Program, pp. 160–165, AAAI, Chicago, 2003.
Download
164.6kB | 75.6kB | 60.4kB |
Abstract
Finding the equilibriated configuration of atomic clusters modeled by the Lennard-Jones potential poses a challenging task to numerical optimization strategies as the number of local minima grows exponentially with the number of atoms in the cluster. We use this massively multimodal problem to test different evolutionary, deterministic and randomized gradient methods with respect to their global search behavior. The randomized gradient method was designed to combine the advantages of gradient and stochastic direct optimization.
BibTeX Entry
@inproceedings{MueSchKou03, author = {Sybille M\"uller and Nicol N. Schraudolph and Petros Koumoutsakos}, title = {\href{http://nic.schraudolph.org/pubs/MueSchKou03.pdf}{ Evolutionary and Gradient-Based Algorithms for {L}ennard-{J}ones Cluster Optimization}}, pages = {160--165}, editor = {Alwyn M. Barry}, booktitle = {Genetic and Evolutionary Computation Conference Workshop Program}, publisher = {AAAI}, address = {Chicago}, year = 2003, b2h_type = {Other}, b2h_topic = {Evolutionary Algorithms}, abstract = { Finding the equilibriated configuration of atomic clusters modeled by the Lennard-Jones potential poses a challenging task to numerical optimization strategies as the number of local minima grows exponentially with the number of atoms in the cluster. We use this massively multimodal problem to test different evolutionary, deterministic and randomized gradient methods with respect to their global search behavior. The randomized gradient method was designed to combine the advantages of gradient and stochastic direct optimization. }}