An Adaptive Penalty-Based Learning Extension for Backpropagation and its Variants

Over the years, many improvements and refinements of the backpropagation learning algorithm have been reported. In this paper, a new adaptive penalty-based learning extension for the backpropagation learning algorithm and its variants is proposed. The new method initially puts pressure on artificial neural networks in order to get all outputs for all training patterns into the correct half of the output range, instead of mainly focusing on minimizing the difference between the target and actual output values. The technique is easy to implement and computationally inexpensive. In this study, the new approach has been applied to the backpropagation learning algorithm as well as the RPROP learning algorithm and simulations have been performed. The superiority of the new proposed method is demonstrated. By applying the extension, the number of successful runs can be greatly increased and the average number of epochs to convergence can be well reduced on various problem instances. Furthermore, the change of the penalty values during training has been studied and its observation shows the active role the penalties play within the learning process.