Evaluation of the computational capabilities of a memristive random network (MN3) under the context of reservoir computing.

This work presents the simulation results of a novel recurrent, memristive neuromorphic architecture, the MN3 and explores its computational capabilities in the performance of a temporal pattern recognition task by considering the principles of the reservoir computing approach. A simple methodology based on the definitions of ordered and chaotic dynamical systems was used to determine the separation and fading memory properties of the architecture. The results show the potential use of this architecture as a reservoir for the on-line processing of time-varying inputs.

Memristive nanowires exhibit small-world connectivity.

Small-world networks provide an excellent balance of efficiency and robustness that is not available with other network topologies. These characteristics are exhibited in the Memristive Nanowire Neural Network (MN3), a novel neuromorphic hardware architecture. This architecture is composed of an electrode array connected by stochastically deposited core-shell nanowires. We simulate the stochastic behavior of the nanowires by making various assumptions on their paths. First, we assume that the nanowires follow straight paths. Next, we assume that they follow arc paths with varying radii. Last, we assume that they follow paths generated by pink noise. For each of the three methods, we present a method to find whether a nanowire passes over an electrode, allowing us to represent the architecture as a bipartite graph. We find that the small-worldness coefficient increases logarithmically and is consistently greater than one, which is indicative of a small-world network.