NEURAL NETWORK WITH THRESHOLD (k, t) STRUCTURE FOR RESIDUAL CODE - BINARY POSITION CODE CONVERSION Russian patent published in 2010 - IPC G06N3/04 

FIELD: information technologies.

SUBSTANCE: invention refers to neural network with threshold (k, t) structure for residual code - binary position code conversion which is the scheme of positional number reconstruction according to abridged system of residual classes modules, no matter that part of modules is distorted and they are discarded or part of modules is simply ignored. For this purpose neural network with threshold (k, t) structure for residual code - binary position code conversion contains input neuron layer, n - neural networks of finite ring to convert residual code into code of generalised positional numbering system, n - constant memory devices to store binary equivalents of generalised positional numbering system coefficients, parallel adder, read-only memory for weight coefficients generation to select necessary configuration of neural networks of finite ring, and read-only memory for output binary code correction generation.

EFFECT: functionality enhancement of residual code - binary position code converter.

1 dwg