A Dynamic Bio-inspired Model of Categorization

Motivated by the outstanding performance of primates in pattern recognition tasks, the main purpose of this research is to exploit the behavioral and neuro-biological findings from primates' visual perception mechanism for categorization applications. Dynamic Bio-Inspired Categorization system (DyBIC) is implemented utilizing nonlinear first order differential equations and its training phase can be accomplished online. The order of the set of differential equations is exclusively a function of the number of categories to be discriminated and the length of the feature vectors doesn't affect system complexity. Besides, the proposed method carries out recognition in a multi-scale mode which is compatible with some of the well-known cognitive and neural phenomena like categorical perception and hierarchical discrimination. The performance of DyBIC is tested on a handmade typical classification example.