1: \begin{abstract}

2: We propose an adaptive algorithm for tracking of historical

3: volatility. The algorithm is built under the assumption that the

4: historical volatility function belongs to the

5: Stone-Ibragimov-Khasminskii class of $k$ times differentiable

6: functions with bounded highest derivative and its subclass of

7: functions satisfying a differential inequalities. We construct an

8: estimator of the Kalman filter type and show optimality of the

9: estimator's convergence rate  to zero as sample size $n\to\infty$.

10: This estimator is in the framework of GARCH design, but

11:  a tuning procedure of its parameters

12: is faster than with traditional GARCH techniques.

13: \end{abstract}

14: