Optimization of logical switching networks in neurostructures

A set of logical functions is found to be implementable using a neurostructure that contains adjoint logical networks. To reduce the number of linear filters, common filtering blocks are marked for a set of neurons of one series. The calculations are based on the Walsh spectral representations. An example of constructing a structure that recognizes three symbols in the 16 × 16 pixel field is given. The factors that ensure the response speed of the neurostructure when implementing it using neurochips are explained.