Method and apparatus for the extraction of a subset of objects by optimizing a measure, using a neural network.

摘要

A neural network includes a stage of construction for a recursive neural network from objects and relations between these objects by associating to each object the binary output of a neuron pi (t) at time t: pi (t) = 1 implies that the object belongs to a searched for sub-assembly; pi (t) = 0 implies that the object does not belong to a searched for sub-assembly. A utilisation stage of an interruption method obtains two-by-two object sub-assemblies not in or in relation and to pass from one sub-assembly to another whilst trying to improve the quality of the solution. An initialisation stage is provided for the output and the potential associated with each neuron. A dynamic recursive running stage is provided for the network. A choosing stage selects the best sub-assemblies listed by comparison with the solution quality.