Salazar, Rafael; Toral, Raul

Hybrid Simulated Annealing using Tsallis Statistics

(4 pages, 2 figures, LaTeX, postscript version here).
We use the Hybrid Monte Carlo method to sample the Tsallis configurational probability p(x) which depends on the parameter q. We use here an homogeneous Markov chain which satisfies the detailed balance condition for all values of q and we apply this method to Simulated Annealing optimization problems. In two examples, a test function with many minima and the discrete Random phase sine-Gordon Model, we show the dependence on the q parameter of the ground state configuration energy obtained after annealing. We conclude that the performance of the simulated annealing method is improved by using q>1 values.

Date published: 1999
Journal: Computer Physics Communications
Volume: 121-122
Page: 40-42

Publications

Nonlinear Science and Statistical Physics Photonics