Developing a Miniature Energy-Harvesting-Powered Edge Device with Multi-Exit Neural Network
Developing a Miniature Energy-Harvesting-Powered Edge Device with Multi-Exit Neural Network
复制标题
DOI：
10.1109/iscas51556.2021.9401799
复制
发表时间：
2021-05
期刊：
影响因子：
--
通讯作者：
Yuyang Li;Yawen Wu;Xincheng Zhang;Ehab A. Hamed;Jingtong Hu;Inhee Lee
中科院分区：
文献类型：
--
作者：
Yuyang Li;Yawen Wu;Xincheng Zhang;Ehab A. Hamed;Jingtong Hu;Inhee Lee
本文介绍了一种微型边缘设备，该设备可以根据可用能量进行不同的出口选项执行神经网络推断。除主路径外，它还提供了一种替代性的早期路径，需要更少的计算，从而增加给定能量的推理操作数量。为了补偿其降级精度，该提议的设备将熵作为早期出口的置信度。该网络使用自定义的低功耗180 nm CMOS处理器芯片和90 nm嵌入式闪存芯片实现，并通过CIFAR-10数据集中的图像进行测试。测量结果表明，与仅Main-Exit方法相比，提出的神经网络将处理时间减少了41.3％，同时牺牲了其准确性从69.5％到66.0％。
This paper describes a miniature edge device that performs neural network inference with different exit options depending on available energy. In addition to the main-exit path, it provides an alternative, early-exit path that requires less computation and thus increase the number of inference operations for given energy. To compensate its degraded accuracy, the proposed device provides entropy as a confidence level for the early exit. The network is implemented with a custom low-power 180 nm CMOS processor chip and a 90 nm embedded flash memory chip and tested by images from CIFAR-10 dataset. The measurement results show the proposed neural network reduces processing time and thus energy consumption by 41.3% compared with the main-exit only method while sacrificing its accuracy from 69.5% to 66.0%.