Towards designing artificial neural networks by evolution
Applied Mathematics and Computation - Special issue on articficial life and robotics
Evolving transfer functions for artificial neural networks
Neural Computing and Applications
Neural network weight training by mutation
Computers and Structures
| Hi-index | 0.00 |
This paper presents the evolutionary training of a feed-forward Artificial Neural Network (ANN) using the Evolutionary Programming (EP). Besides optimizing the ANN regression performance, EP is employed to optimize the architecture and training parameters of a two-hidden layer ANN model for the prediction of total AC power output from a grid-connected photovoltaic system. The Evolutionary Programming-ANN (EPANN) model utilizes solar radiation and ambient temperature as its inputs while the output is the total AC power produced from the grid connected PV system. EP is used to optimize the regression performance of each model by determining the optimum values for the number of nodes in the hidden layer as well as the optimal momentum rate and learning rate for training. The performance of EPANN is tested using ?wo different training algorithms with similar input and output settings. The training algorithms are the Levenberg-Marquardt algorithm and scaled conjugate gradient algorithm. It is found that the Levenberg-Marquardt training algorithm produces better regression performance during training and testing compared to the scaled conjugate gradient training algorithm. Besides that, it could also be implemented faster compared to the scaled conjugate gradient algorithm. Nevertheless, the EPANN with scaled conjugate gradient algorithm could be accomplished using a smaller architecture compared to the EPANN with Levenberg-Marquardt algorithm.