This paper concerns with improving the previously developed RBF equalizer by greatly reducing the number of centers. The basic idea is to select only centers close to the boundary between the different decision classes. The first factor of reducing th...
This paper concerns with improving the previously developed RBF equalizer by greatly reducing the number of centers. The basic idea is to select only centers close to the boundary between the different decision classes. The first factor of reducing the network is ?? where d is the channel delay. The number of centers was further reduced by representing several centers by a single point. This reduction of centers greatly reduces the burden of computation in training, and makes the hardware implementation of RBF equalizers realistic. Simulation studies show that the error rate performance of an RBF equalizer with the proposed reduction in the number of centers compares favorably with the RBF equalizer having the conventional number of centers.