학술논문

A Bayesian approach to parameter selection for simulated annealing
Document Type
Conference
Source
Conference Proceedings 1991 IEEE International Conference on Systems, Man, and Cybernetics Systems, Man, and Cybernetics, 1991. 'Decision Aiding for Complex Systems, Conference Proceedings., 1991 IEEE International Conference on. :613-617 vol.1 1991
Subject
Robotics and Control Systems
Computing and Processing
Bayesian methods
Simulated annealing
Optimization methods
Computational modeling
Convergence
Testing
Computational complexity
Stochastic processes
Piecewise linear approximation
Temperature
Language
Abstract
The authors present a hybrid method of global optimization which uses Bayesian global optimization to choose the parameters of a simulated annealing method, specifically, the initial and final temperatures. This removes the burden on the user for the choice of these values and allows the search to progress optimally. The properties of the Bayesian and simulated annealing algorithms are discussed. A new hybrid algorithm is introduced and selected results on standard test functions are presented.ETX