Guardado en:
Detalles Bibliográficos
Autores principales: You, Haoran, Chen, Xiaohan, Zhang, Yongan, Li, Chaojian, Li, Sicheng, Liu, Zihao, Wang, Zhangyang, Lin, Yingyan Celine
Formato: Preprint
Publicado: 2020
Materias:
Acceso en línea:https://arxiv.org/abs/2010.12785
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866915179959156736
author You, Haoran
Chen, Xiaohan
Zhang, Yongan
Li, Chaojian
Li, Sicheng
Liu, Zihao
Wang, Zhangyang
Lin, Yingyan Celine
author_facet You, Haoran
Chen, Xiaohan
Zhang, Yongan
Li, Chaojian
Li, Sicheng
Liu, Zihao
Wang, Zhangyang
Lin, Yingyan Celine
contents Multiplication (e.g., convolution) is arguably a cornerstone of modern deep neural networks (DNNs). However, intensive multiplications cause expensive resource costs that challenge DNNs' deployment on resource-constrained edge devices, driving several attempts for multiplication-less deep networks. This paper presented ShiftAddNet, whose main inspiration is drawn from a common practice in energy-efficient hardware implementation, that is, multiplication can be instead performed with additions and logical bit-shifts. We leverage this idea to explicitly parameterize deep networks in this way, yielding a new type of deep network that involves only bit-shift and additive weight layers. This hardware-inspired ShiftAddNet immediately leads to both energy-efficient inference and training, without compromising the expressive capacity compared to standard DNNs. The two complementary operation types (bit-shift and add) additionally enable finer-grained control of the model's learning capacity, leading to more flexible trade-off between accuracy and (training) efficiency, as well as improved robustness to quantization and pruning. We conduct extensive experiments and ablation studies, all backed up by our FPGA-based ShiftAddNet implementation and energy measurements. Compared to existing DNNs or other multiplication-less models, ShiftAddNet aggressively reduces over 80% hardware-quantified energy cost of DNNs training and inference, while offering comparable or better accuracies. Codes and pre-trained models are available at https://github.com/RICE-EIC/ShiftAddNet.
format Preprint
id arxiv_https___arxiv_org_abs_2010_12785
institution arXiv
publishDate 2020
record_format arxiv
spellingShingle ShiftAddNet: A Hardware-Inspired Deep Network
You, Haoran
Chen, Xiaohan
Zhang, Yongan
Li, Chaojian
Li, Sicheng
Liu, Zihao
Wang, Zhangyang
Lin, Yingyan Celine
Machine Learning
Multiplication (e.g., convolution) is arguably a cornerstone of modern deep neural networks (DNNs). However, intensive multiplications cause expensive resource costs that challenge DNNs' deployment on resource-constrained edge devices, driving several attempts for multiplication-less deep networks. This paper presented ShiftAddNet, whose main inspiration is drawn from a common practice in energy-efficient hardware implementation, that is, multiplication can be instead performed with additions and logical bit-shifts. We leverage this idea to explicitly parameterize deep networks in this way, yielding a new type of deep network that involves only bit-shift and additive weight layers. This hardware-inspired ShiftAddNet immediately leads to both energy-efficient inference and training, without compromising the expressive capacity compared to standard DNNs. The two complementary operation types (bit-shift and add) additionally enable finer-grained control of the model's learning capacity, leading to more flexible trade-off between accuracy and (training) efficiency, as well as improved robustness to quantization and pruning. We conduct extensive experiments and ablation studies, all backed up by our FPGA-based ShiftAddNet implementation and energy measurements. Compared to existing DNNs or other multiplication-less models, ShiftAddNet aggressively reduces over 80% hardware-quantified energy cost of DNNs training and inference, while offering comparable or better accuracies. Codes and pre-trained models are available at https://github.com/RICE-EIC/ShiftAddNet.
title ShiftAddNet: A Hardware-Inspired Deep Network
topic Machine Learning
url https://arxiv.org/abs/2010.12785