Figure 1

An overview of the GENN method. First, a set of parameters must be initialized in the configuration file. These parameters specify details for the evolutionary processes. Second, the data are divided into 10 equal parts for 10-fold cross-validation. Third, training begins by generating an initial population of random solutions using sensible initialization, which guarantees functional NNs in the initial population. Fourth, each newly generated NN is evaluated on the data in the training set and its fitness recorded. Fifth, a selection technique that is specified by the user is used to select the best solutions for crossover and reproduction in the evolutionary process. The cycle begins with the new generation, which is equal in size to the original population. This cycle continues until either a classification error of zero is found or a limit on the number of generations is reached. After each generation, an optimal solution is identified. At the end of GENN evolution, the overall best solution is selected as the optimal NN. Sixth, this best GENN model is tested on the 1/10 of the data left out to estimate the prediction error of the model. Steps two through six are performed ten times with the same parameters settings, each time using a different 9/10 of the data for training and 1/10 of the data for testing. At the end of a GENN analysis, 10 models are generated – one best model from each cross-validation interval. A final model is chosen based on maximization of the cross-validation consistency of variables/loci across the ten models.