EVOLUTION OF ARTIFICIAL NEURAL-NETWORK ARCHITECTURE - PREDICTION OF DEPRESSION AFTER MANIA

Artificial neural networks (ANNs) are compared to standard statistical methods for outcome prediction in biomedical problems. A general meth od for using genetic algorithms to ''evolve'' ANN architecture (EANN) is presented. Accuracy of logistic regression, a fully interconnected ANN, and an EANN for predicting depression after mania are examined, A ll methods showed very good agreement (training set accuracy, chi-squa re all p <0.01). However, significant differences were found for stabi lity (test set accuracy); logistic regression being the most unstable and EANN being significantly more stable than a fully interconnected A NN (McNemar p <0.01), We conclude that the EANN method enhances ANN st ability. This approach may have particular relevance for biomedical pr ediction problems, such as predicting depression after mania.