Multi-valued neural network and the knowledge acquisition method by the rough sets for ambiguous recognition problem

Proposes a “multi-valued neural network” (MNN) which can distinguish patterns (or failure types) on the basis of the probability distributions of ambiguous feature parameters (or symptom parameters). In most cases of pattern recognition, the knowledge for the recognition are ambiguous. In these cases, the learning data (the diagnosis knowledge) for the MNN must be acquired in some way. Therefore, the knowledge acquisition method for the MNN by using rough sets is also proposed. Several examples of failure diagnosis verify that the methods are effective