No abstract

No abstract

No abstract

No abstract

Multi-valued neurons are the neural processing elements with complex-valued weights, huge functionality (it is possible to implement on the single neuron arbitrary mapping described by partially defined multiple-valued function), quickly converged learning algorithms. Such features of the multi-valued neurons may be used for solution of the dgerent kinds ofproblems. Neural network with multi-valued neurons for image recognition will be considered in the paper. Such a network with original architecture analvzes the phases of the Fourier spectral coeficients corresponding to the low frequencies. Quickly converged learning algorithm and huge functionality of multi-valued neurons allow to get 100% successful recognition of the different classes of images including the blurred and corrupted ones. Simulation results are presented on the example of face recognition.eral papers devoted to Merent types of an associative memory should be mentioned [2-51. The MVN-based cellular neural network has been proposed in [2] as associative memory. The MVN-based neural network with random connections has been proposed [3-41 as associative memory alternative to the Hopfield one. The MVN-based network with random connections requests much smaller number of the connections than fully connected Hopfield network. The quickly converged learning algorithm is another important useful properly of the MVN-based network with random connections. On the other hand Hopfield-like MVN-based neural network has been proposed as associative memory in [5]. A disadvantage of these three networks is impossibility to recognize shifted or rotated images, also as image with changed dynamic range. To break these disadvantages and to use effectively features of the multi-valued neurons, we would like to propose a new type of the network, learning strategy and data representation (frequency domain will be used instead of spatial one).