General Convergence Results for Stochastic Approximations Via Weak Convergence Theory,
Abstract
Using results in the theory of weak convergence of measures and in stability theory for ordinary differential equations, we prove some general convergence theorems for the sequences of random variables which are generated by algorithms of the stochastic approximation type. Such algorithms are used when one wishes to locate, via a recursive Monte-Carlo method, a minimum of a function, under handicap of noisy data. Algorithms for both constrained and unconstrained optimization problems will be considered, and for rather general noise processes. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1976
- Accession Number
- ADA021159
Entities
People
- Harold J. Kushner
Organizations
- Brown University