Alladi Ramakrishnan Hall

Neuroevolution in a small-world dynamical system: Experiments on nanoparticle networks, their computational capacity and simulations based on a mean-field model

Celestine Preetham Lawrence

University of Twente, the Netherlands

This talk begins with a brief introduction to the research field of 'evolution in materio'. It concerns with using matter in its designless form as a computational device - by evolving its external stimuli (e.g. voltages) to provide an appropriate response (e.g. current). We see experimental work on the evolution of a designless nanoparticle cluster into reconfigurable Boolean logic. The nanoparticle cluster is computational because it functions as a single-electron tunnelling (SET) network - which is capable of nonlinearity and inhibition by Coulomb interactions. Next, a novel mean-field approximation is introduced for simulation of SET networks as a nonlinear dynamical system. It converges faster than the traditional Monte Carlo method, with a low trade-off in accuracy. Thus, large-scale small-world nanoparticle networks can now be simulated, more efficiently. We conclude with prospects for 'neuroevolution' in such nanoparticle networks.