Publication IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer SciencesVol.E88-ANo.12pp.3645-3653 Publication Date: 2005/12/01 Online ISSN: DOI: 10.1093/ietfec/e88-a.12.3645 Print ISSN: 0916-8508 Type of Manuscript: PAPER Category: Neural Networks and Bioengineering Keyword: backpropagation, learning, local minima, modified error function,
Full Text: PDF(547.6KB)>>
Summary: One critical "drawback" of the backpropagation algorithm is the local minima problem. We have noted that the local minima problem in the backpropagation algorithm is usually caused by update disharmony between weights connected to the hidden layer and the output layer. To solve this kind of local minima problem, we propose a modified error function with two terms. By adding one term to the conventional error function, the modified error function can harmonize the update of weights connected to the hidden layer and those connected to the output layer. Thus, it can avoid the local minima problem caused by such disharmony. Simulations on some benchmark problems and a real classification task have been performed to test the validity of the modified error function.