Performance analysis of a strategy to distribute and-work and or-work in parallel logic programming systems

Support Vector Machines with Symbolic Interpretation

Inês de Castro Dutra

Haydemar NúñezCecilio AnguloAndreu Catala

This paper studies the performance of Andorra-I, a parallellogic programming system that exploits and-parallelism and or-parallelism with a novel strategy to distribute and-work and or-work among processors. The strategy, work-guided guides it.s decisions by looking at the amount of current and-work and or-work available in an application during execution. The scheduler decision strategy moves workers from one parallel task to another according to the tasks sizes. Results show that the work-guided strategy works quite well and produces better results than the ones produced with a version of Andorra-I that does not allow dynamic migration of workers during execution. We believe that this strategy can be applied to other parallellogic programming systems that aim to exploit both and- and or-parallelism in a single framework. In this work, a procedure for rule extraction from support vector machines is proposed. Our method, first determines prototype vectors by means of k-means. Then, these vectors are combined with the support vectors using geometric methods to define ellipsoids in the input space, which are later translated to if-then rules. In this way, it is possible to give an interpretation to the knowledge acquired by the SVM. On the other hand, the extracted rules render possible the integration of SVMs with symbolic AI systems.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: