A competitive non-linear approach to object recognition: the generalised synergetic algorithm

For many computer vision applications, speed is a fundamental success criterion. The Synergetic Computer using Adjoint Prototypes (SCAP) is a fast linear algorithm which has had several successful industrial implementations. In this work we admit a generalisation of the nonlinear dynamical system on which SCAB is based and show analytically that, for a 2-class classification scheme, the global evolutionary characteristics that allowed a fast linear algorithm to be derived, still hold for this more powerful system. This result is used to develop a new, non iterative training scheme for the generalised system. The training scheme is guaranteed to find an optimum classification of the training data