Title :
Alleviating `overfitting´ via genetically-regularised neural network
Author :
Chan, Z.S.H. ; Ngan, H.W. ; Rad, A.B. ; Ho, T.K.
Author_Institution :
Dept. Of Electr. Eng., Hong Kong Polytech. Univ., Kowloon, China
fDate :
7/18/2002 12:00:00 AM
Abstract :
A hybrid genetic algorithm/scaled conjugate gradient regularisation method is designed to alleviate ANN `over-fitting´. In application to day-ahead load forecasting, the proposed algorithm performs better than early-stopping and Bayesian regularisation, showing promising initial results
Keywords :
conjugate gradient methods; generalisation (artificial intelligence); genetic algorithms; learning (artificial intelligence); load forecasting; neural nets; day-ahead load forecasting; genetically-regularised neural network; hybrid genetic algorithm/scaled conjugate gradient regularisation method; overfitting;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:20020592
