CONTINUOUS-TIME STOCHASTIC APPROXIMATION PROCEDURE WITH RANDOMLY VARYING TRUNCATIONS

In this paper a continuous-time stochastic approximation (S.A.) procedure with randomly varying truncations is proposed, and for its convergence to the value to be sought for it is proved that the conditions imposed on the regression function are rather weak and neither the independence of the increment of the error process nor the boundedness of the procedure are required. The convergence rate and the asymptotic properties for a class of measurement errors are also concerned.