2019 Device Research Conference (DRC) | 2019
Solving the Maximum Independent Set Problem using Coupled Relaxation Oscillators
Abstract
We propose a novel non-Boolean approach that leverages the unique phase dynamics of coupled relaxation oscillators to solve the computationally intractable maximum independent set (MIS; largest sub-graph having no connected vertices) problem in graphs (Fig. 1). Our hardware-centric computing scheme exploits the electrically induced abrupt insulator-metal phase transition (IMT) in VO2, stabilized by a negative feedback, to realize the oscillatory behavior; subsequently, we show that a capacitively coupled system of such oscillators exhibits unique phase dynamics which can be directly mapped to the independent sets (ISs), and thus, the MIS of the network (and topologically equivalent graph). Furthermore, we show the application of this physics-based approach in designing encoding/decoding schemes for data transmission with inherent error correcting capability. Finally, we benchmark our results with existing methods, and show that our hardware enables an efficient approach to solving the NP-Hard MIS problem.