Comment on “Self-organizing neuromorphic nanowire networks as stochastic dynamical systems"

Self-assembled networks of nano-objects have attracted growing interest as physical substrates for brain-inspired computation. The work of Milano et al. 1 is a valuable study of the dynamics of memristive networks of nanowires. The experimental data presented by the authors are comprehensive and of high quality, providing extensive material for validating stochastic conductance models in such systems. The compact Markovian model proposed in ref. 1 , formulated as a stochastic differential equation of the Itô type that combines deterministic return to equilibrium with random components (Gaussian noise and jumps), allows for the description of certain aspects of the dynamics and an assessment of the impact of stochasticity on computational capabilities within the reservoir computing paradigm. However, closer examination of the data from ref. 1 reveals that each key aspect of the conductance dynamics (deterministic relaxation, jump process, and Gaussian fluctuations) exhibits non-Markovian properties, which calls into question the original stochastic model and suggests a need for its fundamental revision.