Thomson-csf

Francois Vallet

Francois Micheron

Jean-Michel Vignolle

Philippe Refregier

Emmanuel De Chambost

Emmanuel Chambost

The disclosure concerns neural networks designed specially for the classification of objects represented by vectors X. If the vectors X include several parameters and if the objects have to be classified in a large number N of classes, the end result is a very large number of interconnections which become difficult to set up physically, are slow in their operation and require lengthy learning phases. The disclosed neural classification system has the particular feature of being constituted on the basis of P neural networks each individually carrying out the classification of objects in only two classes or, at any rate, in a small number of classes only. These networks give probabilities P.sub.i,j of membership in a class C.sub.i among two classes C.sub.i and C.sub.j. The outputs of these networks are connected to a signal processing module which, through simple functions (implementing linear combinations of the outputs and non-linear standardization functions) establishes, on N outputs, results P.sub.i (X) of classification among the N classes. The learning is done on classifications by pairs of classes, but the post-learning recognition gives classifications among N classes.

Neural system of classification and classification method using such a

The neural network of the invention, of the type with a Cartesian matrix, has a first column of addition of the input signals, and at each intersection of the M lines and N columns it comprises a synapse constituted of a simple logic gate.

Programmable analog neural network

The network comprises several memory elements made of ferroelectric polymer, arranged in a matrix organization at the intersections of row and column electrodes. Each memory element (Mij) memorizes a synaptic coefficient a.sub.ij of the network which may be restored by pyroelectric effect on the corresponding column of the network. Amplifier circuits, respectively connected to the columns, give a voltage which is equal to the sum, to which a sign is assigned, of the products of the synaptic coefficients by the voltage components applied to each of the lines of the network.

Growing community of inventors

Montigny Le Bretonneux, France

Francois Vallet