— The idea of using evolutionary techniques to optimize the performance of neural networks is now widely used, but some approaches have been found to result in the evolution of r...
This paper presents FC networks that are instantaneously trained neural networks that allow rapid learning of non-binary data. These networks, which generalize the earlier CC netw...
We propose a dynamically coupled neural oscillator network for image segmentation. Instead of pair-wise coupling, an ensemble of oscillators coupled in a local region is used for ...
The use of monolithic neural networks (such as a multilayer perceptron) has some drawbacks: e.g. slow learning, weight coupling, the black box effect. These can be alleviated by t...
Abstract. A handwritten digit recognition system was used in a demonstration project to visualize artificial neural networks, in particular Kohonen’s self-organizing feature map...