OTS-based Analog-to-Stochastic Converter for Fully-Parallel Weight Update in Cross-point Array Neural Networks

In this study, we experimentally demonstrate a highly linear (R
²
=0.995) and area-efficient 1S1R analog-to-stochastic converter (ASC) which is a core enabler for fully-parallel weight update operation in cross-point synaptic array-based neural networks. We confirm that the previously reported, sigmoid-like ASCs cannot be used for stochastic updates owing to the limitation of nonlinear characteristics. We analyze the characteristics of the device-based ASC for stochastic update operation to show that our ASC consumes 1,000 times less power than a CMOS-based ASC and does not require a power-hungry digital-to-analog converter (DAC). A software-level image recognition accuracy (97.96%) is achieved when performing neural network training with our ovonic threshold switching (OTS) device-based ASC.