医学 AI

BenDi: An Energy-Efficient Quasi-Stochastic Systolic Architecture for Edge Bioelectronics

BenDi: 一种用于边缘生物电子学的节能准随机脉动架构

Bochen Ye, Yihan Pan, Shady Agwa, Themis Prodromakis

专题命中 健康监测 ：diagnosis(abstract)；biomedical(abstract)

AI总结 提出BenDi架构，通过低电压、准随机乘法、脉动数据流和硬件感知量化，在边缘设备上高效运行CNN，相比二进制权重固定脉动架构，面积效率提升3.35倍，能效提升5倍，精度损失仅1%-3.3%。

Comments Accepted for presentation as a short paper at International Conference on Application-specific Systems, Architectures and Processors (ASAP 2026)