Stochastic Comparison of Ellipsoidal and Interval Error Estimation in
Vector Operations
Alexander Ovseevich
We compare accuracy of the ellipsoidal and interval error estimation in the problem of multiplication of an uncertain vector by an exactly known matrix. The advantage of either method is determined by the matrix only. We introduce a model of random matrices and study the probability of advantage of the interval estimate vs. the ellipsoidal one. It is shown that this probability decays as the dimension of uncertain vector goes to infinity.