The compression of the input images in neural network that using method diagonalization the matrices of synaptic weight connections

Abstract

Despite the availability of different algorithms for compression/decompression input images with the emergence of new technical issues of image compression is still relevant, as these tools allow you to assess the capabilities of existing mathematical apparatus that is used for image compression. In addition, the objective of image compression is vital in the design of combined neural networks with a high degree of protection of computer systems against computer attacks DoS, U2L, R2L, Probe and their detection. The decrease in the number of iterations of the random tuning of synaptic connections λik between the i-th and k-th neuron will lead to minimize the setup time of synaptic connections in the neural network and respectively to the quick memorization of information (image). This approach allows to simplify the learning process of the neural network.