Memristive nanowires exhibit small-world connectivity.

Area of Science:

• Neuromorphic Engineering
• Network Science
• Materials Science

Background:

• Small-world networks offer superior efficiency and robustness compared to other topologies.
• Novel hardware architectures are crucial for advancing computational capabilities.

Purpose of the Study:

• To investigate the network characteristics of a Memristive Nanowire Neural Network (MN3).
• To determine if the MN3 architecture exhibits small-world network properties.

Main Methods:

• Simulated stochastic deposition of core-shell nanowires connecting an electrode array.
• Assumed nanowire paths including straight lines, arcs, and pink noise.
• Developed a method to represent the architecture as a bipartite graph by identifying nanowire-electrode intersections.

Main Results:

• The small-worldness coefficient was found to increase logarithmically with varying nanowire path simulations.
• The coefficient consistently exceeded one, indicating small-world network characteristics.
• The MN3 architecture successfully models small-world network properties.

Conclusions:

• The Memristive Nanowire Neural Network (MN3) architecture exhibits small-world network properties.
• This topology provides a balance of efficiency and robustness.
• The simulation methods confirm the potential of MN3 for advanced neuromorphic applications.