@article{oai:miyazaki-u.repo.nii.ac.jp:00002701,
 author = {山森, 一人 and Yamamori, Kunihito and 森元, 聡明 and 吉原, 郁夫 and Yoshihara, Ikuo and Morimoto, Akira},
 journal = {宮崎大学工学部紀要, Memoirs of Faculty of Engineering, University of Miyazaki},
 month = {Aug},
 note = {Some researchers have been proposed to implement neural networks into Wafer Scale Integration(WSI) to achieve fast learning. When neural networks are implemented into a WSI, it has to have a mechanism to avoid hardware defects. To compensate hardware defects, the partial retraining (PR) scheme has proposed. The performance of PR scheme depends on the weights in the neural network because PR scheme only adjusts the weights belonging to a neuron affected by the defects. In this paper, we propose back propagation with annealing scheme (BPA scheme) to improve defect compensation ratio. We show that BPA scheme achieved higher capability of defect compensation than that of conventional BP algorithm.},
 pages = {275--280},
 title = {BPA-PR法による階層型ニューラルネットワークの故障補償能力の向上},
 volume = {37},
 year = {2008},
 yomi = {ヤマモリ, クニヒト and モリモト, アキラ and ヨシハラ, イクオ}
}