NAS-Driven Hardware Accelerator Exploration for Edge AI and Quantization Effects on the Pareto Space
面向边缘AI的NAS驱动硬件加速器探索及量化对帕累托空间的影响
机构 * University of Patras(帕特雷大学)
专题命中 效率与部署 :post-training(abstract);分类 cs.AI
AI总结 本文针对边缘AI部署需求,提出结合NAS、量化与可重构加速器探索的三阶段流水线,实证分析INT4 PTQ对NAS帕累托空间的影响,发现FP32零样本代理覆盖度更优。
Comments 6 pages, 6 figures, accepted for presentation to the 39th IEEE International System-on-Chip Conference, Heidelberg, Germany, September 30 - October 2, 2026