HexCNN: A Framework for Native Hexagonal Convolutional Neural Networks

被引:7
作者
Zhao, Yunxiang [1 ]
Ke, Qiuhong [1 ]
Korn, Flip [2 ]
Qi, Jianzhong [1 ]
Zhang, Rui [1 ]
机构
[1] Univ Melbourne, Melbourne, Vic, Australia
[2] Google Res, Cambridge, MA USA
来源
20TH IEEE INTERNATIONAL CONFERENCE ON DATA MINING (ICDM 2020) | 2020年
关键词
Hexagonal Convolution; Convolutional Neural Networks; Deep Learning;
D O I
10.1109/ICDM50108.2020.00188
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Hexagonal CNN models have shown superior performance in applications such as IACT data analysis and aerial scene classification due to their better rotation symmetry and reduced anisotropy. In order to realize hexagonal processing, existing studies mainly use the ZeroOut method to imitate hexagonal processing, which causes substantial memory and computation overheads. We address this deficiency with a novel native hexagonal CNN framework named HexCNN. HexCNN takes hexagon-shaped input and performs forward and backward propagation on the original form of the input based on hexagon-shaped filters, hence avoiding computation and memory overheads caused by imitation. For applications with rectangle-shaped input but require hexagonal processing, HexCNN can be applied by padding the input into hexagon-shape as preprocessing. In this case, we show that the time and space efficiency of HexCNN still outperforms existing hexagonal CNN methods substantially. Experimental results show that compared with the state-of-the-art models, which imitate hexagonal processing but using rectangle-shaped filters, HexCNN reduces the training time by up to 42.2%. Meanwhile, HexCNN saves the memory space cost by up to 25% and 41.7% for loading the input and performing convolution, respectively.
引用
收藏
页码:1424 / 1429
页数:6
相关论文
共 28 条
[1]  
Abadi M, 2016, ACM SIGPLAN NOTICES, V51, P1, DOI [10.1145/2951913.2976746, 10.1145/3022670.2976746]
[2]  
[Anonymous], 2012, P 2012 INT JOINT C N
[3]  
[Anonymous], 2015, Nature, DOI [DOI 10.1038/NATURE14539, 10.1038/nature14539]
[4]  
[Anonymous], 2017, COMMUN ACM, DOI DOI 10.1145/3065386
[5]  
[Anonymous], 2016, COMPUTER GAMES
[6]  
Chintala S., 2017, 31 C NEURAL INFORM P
[7]   A deep learning-based reconstruction of cosmic ray-induced air showers [J].
Erdmann, M. ;
Glombitza, J. ;
Walz, D. .
ASTROPARTICLE PHYSICS, 2018, 97 :46-53
[8]   Anatomy of high-performance matrix multiplication [J].
Goto, Kazushige ;
Van De Geijn, Robert A. .
ACM TRANSACTIONS ON MATHEMATICAL SOFTWARE, 2008, 34 (03)
[9]  
Hecht-Nielsen R., 1992, Neural Networks for Perception, P65, DOI DOI 10.1016/B978-0-12-741252-8.50010-8
[10]   GEOMETRIC TRANSFORMATIONS ON THE HEXAGONAL GRID [J].
HER, I .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 1995, 4 (09) :1213-1222