Computing with coupled dynamical systems

While Boolean logic has been the backbone of information processing, there are computationally hard problems like optimization and classification wherein this conventional paradigm is less efficient. This motivates us to explore new pathways to their solution. In this invited talk, we describe an experimental testbed using coupled relaxation oscillator arrays that exploit the insulator-metal phase transitions in the correlated material, vanadium dioxide (VO
₂
), to realize a chip scale coupled dynamical system. Our work is inspired by the understanding that associative computing finds a natural analogue in the energy minimization processes embedded within the spatiotemporal dynamics of coupled systems. Our work not only elucidates a physics-based computing method but also presents opportunities for customized co-processors for solving computationally hard problems efficiently.