Adapting DNNs to Embrace Voltage Scaling Induced Errors via Accumulation-Aware Quantization | ArxivCSExplorer