.png){kind=icon}
.png){kind=icon}
Return
Half-Quadratic Criterion based Distributed Adaptive Graph Diffusion Algorithm
.png){kind=icon}
PRE
AI
DOI：10.1016/j.jfranklin.2026.108919.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The vast majority of current works on graph signal processing algorithms assume that the graph structure matrix is an invariant constant matrix, i.e., they only consider static graph signal scenarios. The proposed graph diffusion least mean square (GDLMS) algorithm effectively fills the above gap by successfully transferring the static graph signal scenario to the dynamic graph signal scenario. However, when contaminated by non-Gaussian noise, the performance of the GDLMS algorithm will be seriously degraded, and even the divergence phenomenon occurs. The proposed graph diffusion Generalized Maximum Correntropy Criterion (GDGMCC) algorithm effectively solves the above problem, and it shows better performance against non-Gaussian noise interference. Nevertheless, the non-convex nature of the GMCC function itself limits the convergence/tracking rate and adaptive estimation accuracy of the GDGMCC algorithm in non-Gaussian noise environments. Meanwhile, the GMCC function has three parameters that need to be manually tuned, and the manual parameter tuning process is complex and tedious. To solve the above problems, based on the half-quadratic criterion function with strongly convex property, the graph diffusion half-quadratic criterion (GDHQC) algorithm is proposed and the related performance of the GDHQC algorithm is analyzed in this paper. Finally, the superior performance of the GDHQC algorithm is verified by computer simulation experiments.
Keywords:
Adaptive estimation accuracy
convergence rate
Graph Signal Processing
Generalized Maximum Correntropy Criterion
half-quadratic criterion
Journal
J
IF:
3.7
Papers:
6.3K
Citations:
1.5W
Organization
No organization information available