# A Novel Adaptive Kernel for the RBF Neural Networks

## TL;DR

This paper introduces an adaptive kernel for RBF neural networks that dynamically combines Euclidean and cosine distances, improving performance in estimation tasks like system identification and classification.

## Contribution

It presents a novel adaptive kernel that automatically fuses two distance measures, eliminating the need for manual weight tuning in RBF networks.

## Key findings

- Outperforms manual kernel fusion in experiments
- Effective in nonlinear system identification
- Enhances pattern classification accuracy

## Abstract

In this paper, we propose a novel adaptive kernel for the radial basis function (RBF) neural networks. The proposed kernel adaptively fuses the Euclidean and cosine distance measures to exploit the reciprocating properties of the two. The proposed framework dynamically adapts the weights of the participating kernels using the gradient descent method thereby alleviating the need for predetermined weights. The proposed method is shown to outperform the manual fusion of the kernels on three major problems of estimation namely nonlinear system identification, pattern classification and function approximation.