Knowledge-based neural network: using fuzzy logic to initialize a multilayerd neural network and interpret postlearning results
暂无分享,去创建一个
This paper proposes a neural network system based on fuzzy logic which enables easy conversion between neural networks and fuzzy systems. In the proposed system, the network is initialized based on existing knowledge, and the network extracts knowledge acquired through learning. The system also implements the time-consuming rules of fuzzy systems automatically and tunes membership functions. This paper also formulates a procedure for building structured neural networks based on a fuzzy inference system. The usefulness of the networks was verified by applying them to bond-rating problems. The percentage of correct answers involving unknown data was 96% with our networks, compared to 80% with conventional three-layer neural networks