CONTINUOUS-TIME TWO-STEP STOCHASTIC APPROXIMATION PROCEDURE WITH VARYING TRUNCATIONS AND AN EXAMPLE OF APPLICATION
连续时间变界截尾的两步算法及它的一个应用

A two-step procedure of continuous-time stochastic approximation with randomlyvarying truncations is considered for searching the extremum of a regression function.A priori requiring neither the boundedness of the estimates nor the existence of a Lia-punov function we prove the convergence of the procedure to the value sought for.Aproblem arising in the portfolio theory is illustrated as an example of possible applica-tions.