EMAN : equivalent mass attraction network

This paper introduces a new neural network model for binary pattern classification. In a recent work we have proposed a network model, namely, MAN (Mass Attraction Network) [1,2] which can be used as an autoassociator. In MAN, memory items have been considered as masses at the corners of a hypercube. Exploiting Newton's mass attraction theory, a recall scheme utilizing 'attraction forces' between memory items and input patterns has been developed. EMAN is the consequence of efforts to extent the concept to do classification. The main idea in EMAN is to create an equivalent mass instead of two close masses. After introducing MAN and EMAN concepts, some improvements will be presented. This paper concludes with simulation results.