Deep Neural Network Feasibility Using Analog Spiking Neurons

Abstract

Novel non-Von-Neumann solutions based on artificial intelligence (AI) have surfaced such as the neuromorphic spiking processors in either analog or digital domain. This paper proposes to study the feasibility of deep neural networks on ultra-low-power eNeuron technology. The trade-offs in terms of deep learning capabilities and energy efficiency are highlighted. This study reveals that published eNeurons and synapses satisfy linear fittings for an excitation current greater than 200 pA and a spiking frequency higher than 150 kHz, where energy efficiency is optimal. Thus, deep learning and energy efficiency are mutually exclusive for studied analog spiking neurons.