講演情報
[8a-N306-5]Graph Theory Approach Simulation of Brain-like Silver Nanoparticle Network
〇(D)Tan Thien Dang1, Gianluca Milano3, Yuki Usami1,2, Carlo Ricciardi4, Hirofumi Tanaka1,2 (1.LSSE, Kyutech, 2.Neuromorphic Center, 3.INRiM, 4.PoliTo)
キーワード:
silver nanoparticle network、brain-like network、graph theory
Researchers have been investigating material-based networks as an alternative to traditional crossbar architectures for data processing and computing. To understand these networks, it's crucial to map out their elements and connections, thus creating a structural description of the network architecture. This study focuses on exploring the structural and functional connectivity of self-organized memristive nanoparticle (NP) networks through the application of graph theory.
The study examines how the interconnection topology changes with varying network density distribution using graph metrics. The findings align with percolation theory, demonstrating that the onset of percolation begins in correspondence with the emergence of a giant connected component in the graph induced by a progressive network density increase. The analysis of the ratio of nodes in the largest connected component (NLC/N) relative to network density (D) reveals a sharp transition at the critical density DC ∼ 0.754, where NLC/N corresponds to 1/2. According to percolation theory, this indicates a phase transition from a subcritical state with few edges and many small components to a supercritical state with most nodes connected in a large component. Compared to nanowire networks (DC,NW∼ 5), it suggests NP networks create more connected paths (junctions), leading to a lower critical percolation density, which is considered as a favorable advantage of using NP networks. Additional characterizations show the network exhibit scale-free and small-world properties which provides important computational benefits. These results allow comparison of key network characteristics for a variety of self assembled nanoscale networks and provide a basis for detailed investigations of computational performance.
The study examines how the interconnection topology changes with varying network density distribution using graph metrics. The findings align with percolation theory, demonstrating that the onset of percolation begins in correspondence with the emergence of a giant connected component in the graph induced by a progressive network density increase. The analysis of the ratio of nodes in the largest connected component (NLC/N) relative to network density (D) reveals a sharp transition at the critical density DC ∼ 0.754, where NLC/N corresponds to 1/2. According to percolation theory, this indicates a phase transition from a subcritical state with few edges and many small components to a supercritical state with most nodes connected in a large component. Compared to nanowire networks (DC,NW∼ 5), it suggests NP networks create more connected paths (junctions), leading to a lower critical percolation density, which is considered as a favorable advantage of using NP networks. Additional characterizations show the network exhibit scale-free and small-world properties which provides important computational benefits. These results allow comparison of key network characteristics for a variety of self assembled nanoscale networks and provide a basis for detailed investigations of computational performance.