Visualization of Categorical Data by Hybridization of Two Types of Neural Networks
Masahiro Tanaka* and Hideki Fujiwara**
*Faculty of Science, Konan University 8-9-1 Okamoto, Higashinada-ku, Kobe 658-8501, Japan
**Sakaide Plant, Mitsubishi Chemical Corporation Bannosu-cho, Sakaide, Kagawa 762-8510, Japan
The sandglass neural network is often used for nonlinear auto-association, where the principal information can be extracted by picking up the values of the middle layer. However, the boundary of the classes on this 2-1) surface tends to be complicated because no class information is used. In this paper, the hybridization of auto-associative network and the multi-layer perceptron for classification is proposed. The usefulness of this method is demonstrated by using clinical data.
This article is published under a Creative Commons Attribution-NoDerivatives 4.0 International License.