An method and apparatus for extracting an interpretable, meaningful, and concise rule set from neural networks is presented. The method involves adjustment of gain parameter, .lambda. and the threshold, T.sub.j for the sigmoid activation function of the interactive-or operator used in the extraction/development of a rule set from an artificial neural network. A multi-stage procedure involving coarse and fine adjustment is used in order to constrain the range of the antecedents of the extracted rules to the range of values of the inputs to the artificial neural network. Furthermore, the consequents of the extracted rules are provided based on degree of membership such that they are easily understandable by human beings. The method disclosed may be applied to any pattern recognition task, and is particularly useful in applications such as vehicle occupant sensing and recognition, object recognition, gesture recognition, and facial pattern recognition, among others.

Um método e um instrumento para extrair um jogo interpretable, significativo, e conciso da régua das redes neural são apresentados. O método envolve o ajuste do parâmetro do ganho, o lambda. e o ponto inicial, T.sub.j para a função sigmoid da ativação do interativo-ou operador usado no extraction/development de uma régua ajustada de uma rede neural artificial. Um procedimento multi-stage que envolve o ajuste grosseiro e fino é usado a fim confinar a escala dos antecedentess das réguas extraídas à escala dos valores das entradas à rede neural artificial. Além disso, os conseqüentes das réguas extraídas são fornecidos basearam no grau de sociedade tais que são fàcilmente compreensíveis por seres humanos. O método divulgado pode ser aplicado a toda a tarefa do recognition de teste padrão, e é particularmente útil nas aplicações tais como detetar e recognition do occupant do veículo, recognition do objeto, recognition do gesto, e recognition de teste padrão facial, entre outro.

 
Web www.patentalert.com

< Method and system for predicting transaction

< Learning method and apparatus for a causal network

> Method, system and computer program product for non-linear mapping of multi-dimensional data

> Integral criterion for model training and method of application to targeted marketing optimization

~ 00071