Accession Number : ADA327557

Title :   Dissertation: Autonomous Construction of Multi Layer Perceptron Neural Networks.

Descriptive Note : Final rept. Jul 94-Jun 97,

Corporate Author : AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH SCHOOL OF ENGINEERING

Personal Author(s) : Rathbun, Thomas F.

PDF Url : ADA327557

Report Date : JUN 1997

Pagination or Media Count : 92

Abstract : The construction of Multi Layer Perceptron (MLP) neural networks for classification is explored. A novel algorithm is developed, the MLP Iterative Construction Algorithm (MICA), that designs the network architecture as it trains the weights of the hidden layer nodes. The architecture can be optimized on training set classification accuracy, whereby it always achieves 100% classification accuracies, or it can be optimized for generalization. The test results for MICA compare favorably with those of backpropagation on some data sets and far surpasses backpropagation on others while requiring less FLOPS to train. Feature selection is enhanced by MICA because it affords the opportunity to select a different set of features to separate each pair of classes. The particular saliency metric explored is based on the effective decision boundary analysis, but it is implemented without having to search for the decision boundaries, making it efficient to implement. The same saliency metric is adapted for pruning hidden layer nodes to optimize performance. The feature selection and hidden node pruning techniques are shown to decrease the number of weights in the network architecture from one half to two thirds while maintaining classification accuracy.

Descriptors :   *COMPUTER PROGRAMS, *NEURAL NETS, *NETWORKS, TEST AND EVALUATION, DATA BASES, ALGORITHMS, DECISION MAKING, TRAINING, LAYERS, ACCURACY, NODES, BOUNDARIES, CONSTRUCTION, DECISION THEORY, WEIGHT, CLASSIFICATION, SELF OPERATION, ITERATIONS, MICA.

Subject Categories : Computer Programming and Software

Distribution Statement : APPROVED FOR PUBLIC RELEASE