Title:A Case for Lifetime Reliability-Aware Neuromorphic Computing

Abstract:Neuromorphic computing with non-volatile memory (NVM) can significantly improve performance and lower energy consumption of machine learning tasks implemented using spike-based computations and bio-inspired learning algorithms. High voltages required to operate certain NVMs such as phase-change memory (PCM) can accelerate aging in a neuron's CMOS circuit, thereby reducing the lifetime of neuromorphic hardware. In this work, we evaluate the long-term, i.e., lifetime reliability impact of executing state-of-the-art machine learning tasks on a neuromorphic hardware, considering failure models such as negative bias temperature instability (NBTI) and time-dependent dielectric breakdown (TDDB). Based on such formulation, we show the reliability-performance trade-off obtained due to periodic relaxation of neuromorphic circuits, i.e., a stop-and-go style of neuromorphic computing.