 |
Center for Nonlinear Studies |
We study the agglomeration of conducting particles in between high-voltage electrodes experimentally. We find that the emergent fractal networks are stable and stationary. The degree distribution of the network is highly reproducible: 22% of the particles are endpoints, 22% of the particles are three-fold branching points, and the rest of the particles have degree two. The networks are binary without closed loops. The power consumption decreases during the growth of the network, if the electrical current is kept constant. The response of the networks to an external stimulus is consistent with Hebb's learning rule. Therefore they can be considered as hardware implementations of neural nets. We discuss potential applications in computing and engineering.
Host: Vadas Gintautas