Role of noise in spiking dynamics of diffusive memristor driven by heating-cooling cycles

Recently fabricated and measured diffusive memristors have attracted a significant interest as one of the best candidates to mimic neuron activities and to implement novel computing paradigms. Such devices are capable of exhibiting a combination of dynamical, chaotic, and stochastic phenomena needed for efficient neuromorphic computational systems. However, understanding of contribution and interplay of deterministic and stochastic dynamics to the functional properties of a diffusive memristor is still an open problem. Here, we propose a perturbative approach allowing to separate an influence of noise from regular motion of nanoparticles in diffusive memristors. We demonstrate that noise, coupled with a deterministic degree of freedom in artificial neurons based on diffusive memristors, originates a novel spiking mechanism absent in noiseless systems. The developed formalism suggests an approach towards deterministic modeling of stochastic artificial neurons.