TOPOLOGICAL FEATURE MAPS WITH SELF-ORGANIZED LATERAL CONNECTIONS - A POPULATION-CODED, ONE-LAYER MODEL OF ASSOCIATIVE MEMORY

Guided by the neurobiological principles of self-organization and popu lation coding, we develop a simple, neural, one-layer model for auto-a ssociation. Its core is a feature map endowed with self-organized late ral connections. Input patterns are coded by small spots of active neu rons. The time evolution of neural activity then realizes an auto-asso ciation process by a recurrent attractor dynamics. Population coding i s preserved due to a balance of diffusive spreading of activity and co mpetitive refocusing. Because of its simplicity, the model allows a th orough qualitative and quantitative understanding. We show that the ne twork is capable of performing a cluster analysis and hierarchical cla ssification of data and, thus, qualifies as a tool for unsupervised st atistical data analysis.