Dictionary adaptation for adaptive filtering with multiple kernels in a dynamic environment

One of the major challenges in adaptive filtering is choosing an appropriate model. In this paper, we propose a novel dictionary adaptation rule with multiple kernels. The proposed method can eliminate unnecessary elements in the dictionary by using block ℓ
₁
-regularization. In addition to that, the proposed method can efficiently “move” elements of the dictionary to increase the approximation performance. Numerical examples support the efficacy of the proposed method.