Revised GMDH-Type Neural Networks Using AIC or PSS Criterion and Their Application to Medical Image Recognition
Tadashi Kondo*, Junji Ueno*, and Kazuya Kondo**
*School of Health Sciences, The University of Tokushima, 3-18-15 Kuramoto-cho, Tokushima 770-8509, Japan
**Institute of Health Biosciences, The University of Tokushima, 3-18-15 Kuramoto-cho, Tokushima 770-8503, Japan
This study deals with the revised Group Method of Data Handling (GMDH)-type neural network algorithm using prediction error criterion defined as Prediction Sum of Squares (PSS) or Akaike’s Information Criterion (AIC). The revised GMDH-type neural network algorithm generates optimum multilayered neural network architectures fitting the complexity of nonlinear systems using heuristic self-organization. The revised GMDH-type neural networks self-select the number of layers, optimum neuronal architectures, and useful input variables to minimize prediction error criterion defined as PSS or AIC. This algorithm is applied to the identification problem of the nonlinear complex system and results are compared to those obtained by the revised GMDH algorithm and conventional multilayered neural networks. The revised GMDH-type neural network algorithm is also applied to medical image recognition and it is shown that this algorithm is useful for medical image recognition.
This article is published under a Creative Commons Attribution-NoDerivatives 4.0 International License.