
Adaptive importance sampling of random walks on continuous state spaces
Author(s) -
Keith A. Baggerly,
Dennis D. Cox,
Richard R. Picard
Publication year - 1998
Language(s) - English
Resource type - Reports
DOI - 10.2172/677157
Subject(s) - random walk , mathematics , sampling (signal processing) , convergence (economics) , exponential function , variance (accounting) , state (computer science) , state space , finite state , space (punctuation) , statistical physics , calculus (dental) , statistics , computer science , algorithm , mathematical analysis , markov chain , physics , medicine , accounting , dentistry , filter (signal processing) , economics , computer vision , economic growth , business , operating system
The authors consider adaptive importance sampling for a random walk with scoring in a general state space. Conditions under which exponential convergence occurs to the zero-variance solution are reviewed. These results generalize previous work for finite, discrete state spaces in Kollman (1993) and in Kollman, Baggerly, Cox, and Picard (1996). This paper is intended for nonstatisticians and includes considerable explanatory material