Electrochemical Ionic Synapses for Analog Deep Learning and Beyond

Deep learning is a powerful algorithm for machine learning applications such as computer vision and natural language processing However, the training of these neural networks is limited by the traditional von Neumann architecture of our current CPUs and GPUs, which results in significant energy consumption In this talk, I will share our work on the ionic electrochemical synapses, whose conductivity we can control deterministically by electrochemical insertion/extraction of dopant ions across the active device layer The protons present very low energy consumption, on par with biological synapses in the brain, while magnesium ions present with better stability without the need for encapsulation The modeling results indicate the desirable material properties, such as ion conductivity and interface charge transfer kinetics, that we must achieve for fast ( low energy fJ and low voltage 1 V) performance of these devices Our findings provide pathways towards brain inspired hardware that has high yield and consistency and uses significantly lesser energy as compared to current computing architectures