Polymer-Linked Nanoparticle Networks Running on Heat Can Act as Computing Devices

Abstract Developing physical neural network (PNN) hardwares is important to next-generation artificial intelligence systems. Phononic devices─using heat current to encode and process information─is one of the solutions to neuromorphic computing. In this work, we backmap an artificial neural network (ANN) into a PNN simulation model using polymer networked nanoparticles (PNNPs). Our atomistic simulation results demonstrate that the polymer-linked nanoparticle networks can potentially realize information processing─viz., classification of input temperatures according to the output heat currents. Using high-throughput molecular dynamics (MD) simulations and the trust region Bayesian optimization (TuRBO) methods, we tune the plasticity of polymer linkers and the temperatures of nanoparticles─28 parameters in total─to optimize the performance of the PNNP machines, which is analogous to tuning of the weights and bias of the corresponding ANNs. After five rounds of high-throughput MD simulations, we demonstrate that the PNNP machine can better label different input temperatures, and thus its performance is improved. We also use a testing data set to verify the heat flow outputs from the top five PNNP machines in each round.