The Combinatorial Neural Model (CNM) ([8] and [9]) is a hybrid architecture for intelligent systems that integrates symbolic and connectionist computational paradigms. This model has shown to be a good alternative to be used on data mining; in this sense some works have been presented in order to deal with scalability of the core algorithm to large databases ([2,1] and [10]). Another important issue is the prunning of the network, after the trainingp hase. In the original proposal this prunningi s done on the basis of accumulators values. However, this criterion does not give a precise notion of the classification accuracy that results after the prunning. In this paper we present an implementation of the CNM with a feature based on the wrapper method ([6] and [12]) to prune the network by usingt he accuracy level, instead of the value of accumulators as in the original approach.
CITATION STYLE
Prado, H. A., Machado, K. F., Frigeri, S. R., & Engel, P. M. (1999). Accuracy tuning on combinatorial neural model. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1574, pp. 247–252). Springer Verlag. https://doi.org/10.1007/3-540-48912-6_33
Mendeley helps you to discover research relevant for your work.