Memristor models for pattern recognition systems

Abstract

The design of Memristor Oscillatory Neurocomputers for pattern recognition tasks may not leave aside a preliminary thorough investigation of the nonlinear dynamics of the whole system and its basic components. This chapter yields novel insights into the peculiar nonlinear dynamics of different memristor models.A detailed mathematical treatment aimed at highlighting the key impact the initial condition on the flux across a memristor with odd-symmetric charge-flux characteristic has on the development of a particular dynamical behavior. It is proved how, driving the memristor with a sine-wave voltage source, the amplitude-angular frequency ratio selects a sub-class of observable current-voltage behaviors from the class of all possible dynamics, while the initial condition on flux specifies which of the behaviors in the sub-class is actually observed. In addition, a novel boundary condition-based model for memristor nano-scale films points out how specification of suitable dynamical behavior at film ends, depending on the particular physical realization under study and on driving conditions, crucially determines the observed dynamics.