Title :
Multiobjective Evolutionary Optimization of Training and Topology of Recurrent Neural Networks for Time-Series Prediction
Author :
Katagiri, H. ; Nishizaki, I. ; Hayashida, T. ; Kadoma, T.
Author_Institution :
Grad. Sch. of Eng., Hiroshima Univ., Hiroshima, Japan
Abstract :
This paper provides a new evolutionary multiobjective optimization method for automatically optimizing the network topology of recurrent neural networks. The feature of the proposed method is that it involves a procedure of intensively exploring a region including solutions with small training errors in the Pareto frontier, instead of finding a whole set of the Pareto optimal solutions. Several numerical experiments are executed in order to show the advantage of the proposed method over the existing effective algorithm by Delgado et al. with respect to the capability of time-series prediction.
Keywords :
Pareto analysis; evolutionary computation; numerical analysis; recurrent neural nets; Pareto optimal solutions; multiobjective evolutionary optimization; network topology; recurrent neural networks; time-series prediction; training; Character recognition; Computational efficiency; Computer networks; Handwriting recognition; Inference algorithms; Network topology; Neural networks; Optimization methods; Recurrent neural networks; Wind speed;
Conference_Titel :
Information Science and Applications (ICISA), 2010 International Conference on
Conference_Location :
Seoul
Print_ISBN :
978-1-4244-5941-4
DOI :
10.1109/ICISA.2010.5480391