Related papers: Cross-Layer Optimization for Fault-Tolerant Deep Learning

Cross-Layer Optimization for Fault-Tolerant Deep Learning

URL: http://arxiv.org/abs/2312.13754v1
Date: Thu, 21 Dec 2023 11:35:45 GMT
Title: Cross-Layer Optimization for Fault-Tolerant Deep Learning
Authors: Qing Zhang, Cheng Liu, Bo Liu, Haitong Huang, Ying Wang, Huawei Li, Xiaowei Li
Abstract summary: We propose to characterize deep learning vulnerability difference across both neurons and bits of each neuron, and leverage the vulnerability difference to enable selective protection of the deep learning processing components. We employ Bayesian optimization strategy to co-optimize the correlated cross-layer design parameters at algorithm layer, architecture layer, and circuit layer.
Score: 17.724727744611535
License: http://creativecommons.org/licenses/by/4.0/
Abstract: Fault-tolerant deep learning accelerator is the basis for highly reliable deep learning processing and critical to deploy deep learning in safety-critical applications such as avionics and robotics. Since deep learning is known to be computing- and memory-intensive, traditional fault-tolerant approaches based on redundant computing will incur substantial overhead including power consumption and chip area. To this end, we propose to characterize deep learning vulnerability difference across both neurons and bits of each neuron, and leverage the vulnerability difference to enable selective protection of the deep learning processing components from the perspective of architecture layer and circuit layer respectively. At the same time, we observe the correlation between model quantization and bit protection overhead of the underlying processing elements of deep learning accelerators, and propose to reduce the bit protection overhead by adding additional quantization constrain without compromising the model accuracy. Finally, we employ Bayesian optimization strategy to co-optimize the correlated cross-layer design parameters at algorithm layer, architecture layer, and circuit layer to minimize the hardware resource consumption while fulfilling multiple user constraints including reliability, accuracy, and performance of the deep learning processing at the same time.