Nonlinear Ion Dynamics Enable Spike Timing Dependent Plasticity of Electrochemical Ionic Synapses

Abstract

Programmable synaptic devices that can achieve timing-dependent weightupdates are key components to implementing energy-efficient spiking neuralnetworks (SNNs). Electrochemical ionic synapses (EIS) enable theprogramming of weight updates with very low energy consumption and lowvariability. Here, the strongly nonlinear kinetics of EIS, arising from nonlineardynamics of ions and charge transfer reactions in solids, are leveraged toimplement various forms of spike-timing-dependent plasticity (STDP). Inparticular, protons are used as the working ion. Different forms of the STDPfunction are deterministically predicted and emulated by a linearsuperposition of appropriately designed pre- and post-synaptic neuronsignals. Heterogeneous STDP is also demonstrated within the array tocapture different learning rules in the same system. STDP timescales arecontrollable, ranging from milliseconds to nanoseconds. The STDP resultingfrom EIS has lower variability than other hardware STDP implementations,due to the deterministic and uniform insertion of charge in the tunablechannel material. The results indicate that the ion and charge transferdynamics in EIS can enable bio-plausible synapses for SNN hardware withhigh energy efficiency, reliability, and throughput.