Estimating Network Reliability with Accelerated Convergence Rates on Error Bounds

摘要

The purpose of this paper is to describe related estimation techniques that improve the convergence rate to 1/K when a specified nonrandom finite sequence of sample points is used and to (log K)/K when certain nonrandom infinite sequences of sample points are used. In addition, algorithms are given for computing estimates of g(s,t) with 1/K and (log K)/K convergence rates with computation time complexities per replication having an upper bound O(N) + O(max(N,/V/)) as K approaches limit of infinity. Although the infinite sequences have a slower convergence rate than the finite sequence has, these offer the convenience of allowing one to add replications as desired and retaining the rate. By contrast, the finite sequence does not allow this addition, once K is initially fixed.