Exponentially convergent behaviour of simple stochastic adaptive estimation algorithms
Robert R. Bitmead, Brian D. O. Anderson · 1978
A stochastic algorithm, familiar from adaptive estimation, is introduced and its homogeneous part is shown to be exponentially convergent for a wide class of inputs, which need not be stationary. The implications of this convergence rate for the nonhomogeneous algorithm in practical situations are qualitatively examined and a possible approach to improving performance in use is suggested.