|کد مقاله||کد نشریه||سال انتشار||مقاله انگلیسی||ترجمه فارسی||نسخه تمام متن|
|4977394||1451925||2018||8 صفحه PDF||سفارش دهید||دانلود کنید|
- The LLAD algorithm is firstly introduced into the adaptive network, where nodes are equipped with LLAD cost function. We have obtained the global cost function over adaptive network.
- In order to develop the diffusion LLAD algorithm, we approximate the global cost function by local cost functions which are amenable to distributed optimization. The diffusion LLAD is based on local cost function.
- Compare with centralized LLAD algorithm, the diffusion LLAD has a good balance between performance and communications.
- The diffusion LLAD performs better than diffusion LMS and diffusion sign-error LMS in impulsive noise environment.
The popular distributed estimation algorithms based on the mean-square error criterion is not robust against impulsive noise in the adaptive networks. To address the problem, the diffusion least logarithmic absolute difference (LLAD) algorithm is proposed in this article, which adopts both the logarithm operation and sign operation to the error. The algorithm can elegantly and gradually adjust the conventional cost functions in its optimization based on the error variation. Compared with centralized LLAD algorithm, the diffusion LLAD algorithm performs a good balance between communications and performance. The theoretical stability of mean and mean-square performance of the algorithm is analyzed. Simulation results indicate that the algorithm achieves a better performance, compared with diffusion LMS and diffusion sign-error LMS algorithms, even in the impulsive noise environment.
Journal: Signal Processing - Volume 142, January 2018, Pages 423-430