Dynamic Task Remapping for Reliable CNN Training on ReRAM Crossbars
Dynamic Task Remapping for Reliable CNN Training on ReRAM Crossbars
复制标题
DOI：
10.23919/date56975.2023.10137238
复制
发表时间：
2023-04
期刊：
影响因子：
--
通讯作者：
C. Tung;B. K. Joardar;P. Pande;J. Doppa;Hai Helen Li;K. Chakrabarty
中科院分区：
文献类型：
--
作者：
C. Tung;B. K. Joardar;P. Pande;J. Doppa;Hai Helen Li;K. Chakrabarty
基于ReRAM交叉杆的计算系统（RCS）可以加速CNN训练。然而，由于制造缺陷和有限的耐久性导致的硬件故障阻碍了RCS的广泛采用。我们提出了一种基于动态任务重映射的技术，用于在有故障的RCS上进行可靠的CNN训练。实验结果表明，在存在故障的情况下，使用CIFAR-IO、CIFAR-100和SVHN数据集训练流行的CNN（如VGG、ResNets和SqueezeNet）时，所提出的低开销方法平均仅会导致0.85%的准确率损失。
A ReRAM crossbar-based computing system (RCS) can accelerate CNN training. However, hardware faults due to manufacturing defects and limited endurance impede the widespread adoption of RCS. We propose a dynamic task remapping-based technique for reliable CNN training on faulty RCS. Experimental results demonstrate that the proposed low-overhead method incurs only 0.85% accuracy loss on average while training popular CNNs such as VGGs, ResNets, and SqueezeNet with the CIFAR-IO, CIFAR-100, and SVHN datasets in the presence of faults.