Graphlet Decomposition Using Random-Walk Memristors

Although memristor crossbars are a promising post-CMOS solution for computing, sneak currents and stochastic switching are two persistent challenges that impede their practical implementation. Here, we show how both issues can, in fact, be taken advantage for energy efficient computing. Using sneak paths to represent graphlets and stochasticity in hybrid volatile-nonvolatile memristors to mimic random walks, we perform graphlet decomposition and analysis, which are computationally hard problem with various applications, such as social networking and genome slicing.