Learning algorithm and neurocomputing architecture for NDS Neurons

We implement a learning algorithm for Nonlinear Dynamic State (NDS) Neurons in the framework of Nonlinear Transient Computation (NTC). The learning procedure is based on Spike-Timing Dependent Plasticity (STDP); which maintains the nonlinear dynamics of these neurons so they can perform classification of time varying signals. To expound the practicality of this approach, an example of forgery detection for Online Signature Verification is presented. Also, we speculate on the importance of the presented work in modelling basic cognitive processes (e.g. memory) and its relation to chaotic neurodynamics.