Metropolis-type Annealing Algorithms for Global Optimization in IRd
Abstract
We establish the convergence of a class of Metropolis-type Markov chain annealing algorithms for global optimization of a smooth function U(.) on IRd. No prior information is assumed as to what bounded region contains a global minimum. Our analysis is based on writing the Metropolis-type algorithm in the form of a recursive stochastic algorithm, where [some entities] are independent standard Gaussian random variables, [and others] are (unbounded, correlated) random variables, and then applying results about [our findings].
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1990
- Accession Number
- ADA459610
Entities
People
- Sanjoy K. Mitter
- Saul B. Gelfand
Organizations
- Massachusetts Institute of Technology