ニューロン回路 : パターン認識(<小特集>ニューロコンピューティング)

概要

論文の詳細を見る
Many artificial neural network models and learning algorithms have been developed and applied to pattern recognition. In this paper, pattern recognition methods using artificial neural network models are supervised and categorized according to how neural network are used. In most of the methods, both the feature extraction and the classification are implemented with neural network technique, but the performance is not necessarily higher than that of conventional pattern recognition methods. Some neural networks used in other processing steps of pattern recognition are also explained. Then some problems of applying artificial neural networks to pattern recognition are studied, such as the difficulty of learning control, the long learning time, a huge number of required sample patterns, and incapability of invariant feature extraction. One of the solutions is to construct inherent network layers, which can extract invariant features, in a neural networks. Some other ideas are discussed for constructing artificial neural networks enabling higher performance pattern recognition.
日本シミュレーション学会の論文
1989-03-15