Many people expect the third function of foods called as physiological activities that affect our health condition. However, measurement of a physiological activity is troublesome and measurements for all kinds of foods are not impractical. Therefore a system which can easily estimate physiological activities is required. We have proposed a method to estimate physiological activities of foods from protein expression levels using artificial neural networks (ANNs). Estimation of physiological activity using conventional ANN has problems that physiological activities take positive real number more than 1.0, but dynamic range of ANN is limited from 0.0 to 1.0. And estimation accuracy was not enough. To solve those problems, we employ multi-modal neural network with non-threshold activation function. By using multi-modal neural network with non-threshold activation function, ANN can directly handle physiological activities as the training signals those are more than 1.0. Experimental results showed that our method improved estimation accuracy than that of conventional ANN with scaled training samples for some physiological activities.
雑誌名
宮崎大學工學部紀要
巻
40
ページ
239 - 244
発行年
2011-07-30
出版者
宮崎大学工学部
Faculty of Engineering, University of Miyazaki
リポジトリのご案内
非しきい値型活性化関数を持つ神経回路網群による生理活性推定法
41engineering40_pp.239-244.pdf
(793.66KB)
