Using Multiple Self-Organizing Architectures for Object Recognition Using Multiple Self-Organizing Architectures for Object Recognition

Abstract- This work proposes a model of the development of visual object recognition, based on the combination of two dierent articial neural architectures, both supporting self-organization: LISSOM and SOM. The former is a better approximation of the biological computations in cortical areas, including lateral connections, the latter is best suited for a simple synthesis of non localized processes, like object categorization. Key words- object recognition, self-organizing maps, visual invariance