1: \begin{abstract}

2: We introduce a new analysis of an adaptive mixture method that

3: combines outputs of two constituent filters running in parallel to

4: model an unknown desired signal.  This adaptive mixture is shown to

5: achieve the mean square error (MSE) performance of the best

6: constituent filter, and in some cases outperforms both, in the

7: steady-state. However, the MSE analysis of this mixture in the

8: steady-state and during the transient regions uses approximations and

9: relies on statistical models on the underlying signals and

10: systems. Hence, such an analysis may not be useful or valid for

11: signals generated by various real life systems that show high degrees

12: of nonstationarity, limit cycles and, in many cases, that are even

13: chaotic. To this end, we perform the transient and the steady-state

14: analysis of this adaptive mixture in a ``strong'' deterministic sense

15: without any approximations in the derivations or statistical

16: assumptions on the underlying signals such that our results are

17: guaranteed to hold. In particular, we relate the time-accumulated

18: squared estimation error of this adaptive mixture at any time to the

19: time-accumulated squared estimation error of the optimal convex

20: mixture of the constituent filters directly tuned to the

21: underlying signal in an individual sequence manner.

22: %We also demonstrate that lowering the learning rate improves the final performance at the cost of a low convergence under the deterministic setting.

23: %We investigate a convexly constrained mixture method, Arenas-Garcia

24: %et. al, 2006, that  outputs of two constituent

25: %filters running in parallel on the same task. We study this adaptive

26: %mixture method under a deterministic setting, where, unlike in Arenas-Garcia

27: %et. al, 2006, we refrain from making stochastic assumptions on the underlying signals

28: %or outputs such that our results are guaranteed to hold in a ``strong''

29: %deterministic sense.

30: \end{abstract}

31: