A MYKOLAIV neuron network is intended for associative recognition of patterns described by vectors with elements ±1, has input first sensory layer with n neurons intended for multiplication of input signal X=(x1, x2, …, xj, …, xn), each neuron having one input with weighting factor 1 and m outputs, each connected to corresponding input of second associative layer neuron. Each neuron of the second associative layer in amount of m has n inputs and one output, and weighting factors of n inputs of each neuron of the second associative layer are equal to corresponding elements of reference pattern XEi=(xEi, x2Ei, …, xjEi, …, xnEi, wherein i = 1, 2, ..., m is a sequential number of reference standard, and output of each of m neurons of the second associative layer is intended for definition of scalar product of two vectors as zi={(X)T[1:n]·(XEi)[n:1]}, m outputs of neurons of the second associative layer are intended for signal transmission as vector Z=(z1, z2, …, zi, …, zm) and connected to m inputs with weighting factors 1 of responsive layer having m outputs for derivation of output vector Y=(y1, y2, …, yi, …, ym), each element yj of which is intended for derivation of 0 value if corresponding value of element zi of vector Z=(z1, z2, …, zi, …, zm) is less than maximum value of Z vector elements.
展开▼
