1: \begin{abstract}    

2: 	A simple yet effective architectural design of radial basis function neural networks (RBFNN) makes them amongst the most popular conventional neural networks. The current generation of radial basis function neural network is equipped with multiple kernels which provide significant performance benefits compared to the previous generation using only a single kernel. In existing multi-kernel RBF algorithms, multi-kernel is formed by the convex combination of the base/primary kernels. In this paper, we propose a novel multi-kernel RBFNN in which every base kernel has its own (local) weight. This novel flexibility in the network provides better performance such as faster convergence rate, better local minima and resilience against stucking in poor local minima. These performance gains are achieved at a competitive computational complexity compared to the contemporary  multi-kernel RBF algorithms. The proposed algorithm is thoroughly analysed for performance gain using mathematical and graphical illustrations and also evaluated on three different types of problems namely: 

3: 		\begin{inparaenum}[(i)]

4: 		\item pattern classification, 

5: 		\item system identification and 

6: 		\item function approximation.

7: 		\end{inparaenum}

8: 	Empirical results clearly show the superiority of the proposed algorithm compared to the existing state-of-the-art multi-kernel approaches.

9: %	(1) pattern classification, system identification and function approximation.

10: 	\keywords{{pattern classification} \and {function approximation} \and non-linear system identification \and {neural networks} \and {radial basis function} \and {Gaussian kernel} \and {support vector machine} \and {euclidean distance} \and {cosine distance} \and {kernel fusion}}

11: 	% \PACS{PACS code1 \and PACS code2 \and more}

12: 	% \subclass{MSC code1 \and MSC code2 \and more}

13: \end{abstract}

14: