Investigating better multi-layer perceptrons for the task of classification
Hyontai Sug · WSEAS Transactions on Computers archive · 2010
The task of deciding proper sample sizes for multi-layer perceptrons tends to be arbitrary so that, depending on sample data sets, the performance of trained multi-layer perceptrons has a tendency of some fluctuation. As sample size grows, multi-layer perceptrons have the property that performance in prediction accuracy becomes better slowly with some fluctuation. In order to exploit this property this paper suggests a progressive and repeated sampling technique for better multi-layer perceptrons to cope with the fluctuation of prediction accuracy that depend on samples as well as the size of samples. Experiments with six different data sets in UCI machine learning repository showed very good results.