Logic Neuron Model Using Walsh Spectral Representations

摘要

The weak effectiveness is shown of the weight setting for the neurons with a large number of inputs. A model is proposed of a neuron with increased functionality. The model is based on the frequency principle of weight setting. The interaction of neurons in the retina is analyzed and a model is proposed of the corresponding sections of the neural network, which includes a neuron and a block of linear filters. The corresponding devices are developed on the basis of spectral representations of logic functions (LFs) according to Walsh. The rationale is given for the spectral analysis of families partially defined by LFs. Suppositions are made about the role of filtering in brain activity during recognition of images.