This paper first introduces the basic knowledge and the current development of neural networks, fuzzy sets,and rough sets. Chapter Two is mainly about reduction algorithm based on the attributes of dependability. This algorithm is in fact reduced on the basis of the knowledge of rough sets, and it is obviously improved comparing with the former ones. Chapter Three and Chapter Four discuss respectively how to constitute rough neural networks and fuzzy neural networks with the knowledge of rough sets. Each of these networks has their own advantages when applied to solve mathematic problems. In Chapter Five, the author combines the advantages of rough neural networks and fuzzy neural networks, and constitutes a new neural network, which is simpler and has a better function. The paper ends with the questions put forward by the author which deserve further research. |