A new approach for finding the weights in neural network using graphs

This paper describes a method for obtaining the weights of a Hopfield network exploring the possibility of using graph theory. We associate each pattern going to be stored in the network with a certain basic graph and as a function of these basic graphs, we obtain a resulting graph. This graph contains the weights to be assigned to the network connections. A measure function may be defined on this graph, which can be used to classify the new patterns presented to the network during the testing process. The aim of the present research is intended to explore new specification techniques of neural networks based on graphs to be used in the optimization and simplification of network architectures and computational complexity. This methodology can be used to treat real problems in pattern recognition and speech processing