21 April 2021 in IEEE Transactions on Circuits and Systems II: Express Briefs
In this brief, a kernel adaptive filter based on the Student’s ${t}$ distribution in the reproducing kernel Hilbert space (RKHS) is presented, which is distinct from the traditional kernel adaptive filtering algorithms as follows: first, a Student’s ${t}$ reproducing kernel function is proposed to fight against the abrupt noise together with Gaussian noise depicted by the impulsive-Gaussian mixed noise model; and second, a Strengthened Surprise Criterion (SSC) is devised to reduce the size of the neural networks, which is utilized to implement the proposed Student’s ${t}$ -based kernel filter. The proposed algorithms are compared with the widely used KLMS and recently proposed KRLS-type filters in terms of the accuracy error under both Gaussian and abrupt noise. Experimental results show that the proposed Student’s ${t}$ -based kernel adaptive filter can improve the estimation accuracy at least by 20% while having more compact size of neural networks compared with the existed kernel adaptive algorithms.
23 November 2022 in IEEE Transactions on Circuits and Systems II: Express Briefs
The multikernel adaptive filters (MKAFs) have been successfully applied to resolve the issue of kernel parameter selection in traditional single kernel adaptive filters. However, owing to the linear growing network structures, conventional MKAFs commonly suffer a lot from large computational and memory burdens. To solve this problem, a multiple Nyström approximation is proposed to curb the computational complexity of MKAFs in this brief. More concretely, the multiple Nyström method is incorporated into the kernel generalized Cauchy conjugate gradient algorithm, generating a novel multiple Nyström kernel generalized Cauchy conjugate gradient algorithm (MNKGCCG). It is noted that the MNKGCCG can achieve the desirable filtering performance with low computational cost in the fixed-dimensional feature space. Experimental results on Mackey-Glass time series and sunspots time series predictions in non-Gaussian noise environments demonstrate the superiorities of the proposed MNKGCCG algorithm in terms of filtering accuracy and robustness.