RankSearch: An Automatic Rank Search Towards Optimal Tensor Compression for Video LSTM Networks on Edge

Changhai Man1; Cheng Chang2; Chenchen Ding3; Ao Shen3; Hongwei Ren3; Ziyi Guan4; Yuan Cheng5; Shaobo Luo3; Rumin Zhang3; Ngai Wong4; Hao Yu3

10.23919/DATE56975.2023.10137115

2023

Design, Automation and Test in Europe Conference and Exhibition (DATE)

1530-1591

979-8-3503-9624-9

2023 Design, Automation & Test in Europe Conference & Exhibition (DATE)

2023-April

1-2

17-19 April 2023

Antwerp, Belgium

345 E 47TH ST, NEW YORK, NY 10017 USA

IEEE

Various industrial and domestic applications call for optimized lightweight video LSTM network models on edge. The recent tensor-train method can transform space-time features into tensors, which can be further decomposed into low-rank network models for lightweight video analysis on edge. The rank selection of tensor is however manually performed with no optimization. This paper formulates a rank search algorithm to automatically decide tensor ranks with consideration of the trade-off between network accuracy and complexity. A fast rank search method, called RankSearch, is developed to find optimized low-rank video LSTM network models on edge. Results from experiments show that RankSearch achieves a 4.84x reduction in model complexity, and 1.96x speed-up in run time while delivering a 3.86% accuracy improvement compared with the manual-ranked models.

Tensor Rank Search Complexity Aware Optimization Tensorized Video LSTM Edge Computing

其他

英语

CPCI-S ; EI

Automation & Control Systems ; Computer Science ; Engineering

Automation & Control Systems ; Computer Science, Hardware & Architecture ; Engineering, Industrial

WOS:001027444200163

20232614287775

Complex networks ; Economic and social effects ; Edge computing ; Long short-term memory

Computer Systems and Equipment:722 ; Digital Computers and Systems:722.4 ; Algebra:921.1 ; Social Sciences:971

IEEE

1.Georgia Institute of Technology
2.University of California, Los Angeles
3.Southern University of Science and Technology
4.The University of Hong Kong
5.Shanghai Jiao Tong University

Changhai Man,Cheng Chang,Chenchen Ding,et al. RankSearch: An Automatic Rank Search Towards Optimal Tensor Compression for Video LSTM Networks on Edge[C]. 345 E 47TH ST, NEW YORK, NY 10017 USA:IEEE,2023:1-2.