A Multilayer Neural Accelerator With Binary Activations Based on Phase-Change Memory

Novel in-memory computing circuits, based on arrays of emerging nonvolatile memories, such as the phase-change memory (PCM), can boost cutting-edge performances artificial intelligent applications. However, spread PCM-based circuits is currently hindered by lack a design framework enabling fast, efficient, and low-power neural networks. In this work, novel approach to conceptual technical integrated networks proposed. particular, relax power hunger complexity state-of-the-art solutions, we propose fully analog where analog-to-digital converter (ADC) replaced simple comparator. The building blocks accelerator are presented validated in Cadence Virtuoso. major nonidealities, PCM conductance variability, drift, IR drop, readout threshold, studied considering their impact accuracy.