Frequency-Domain Acceleration for 3D Generative Adversarial Networks

被引：0
|
作者：
Jiang Z. [1 ]
Liu B. [1 ,2 ]
Tang Y. [3 ]
Wu J. [1 ]
机构：
[1] School of Computer Science and Technology, Guangdong University of Technology, Guangzhou
[2] State Key Laboratory of Computer Architecture, Institute of Computing Technology, Chinese Academy of Sciences, Beijing
[3] Wuhan Digital Engineering Institute, Wuhan
来源：
Jisuanji Fuzhu Sheji Yu Tuxingxue Xuebao/Journal of Computer-Aided Design and Computer Graphics | 2023年 / 35卷 / 06期
关键词：
3D generative adversarial networks; data reuse; deconvolution; frequency-domain accelerator; stream scheduling;
D O I：
10.3724/SP.J.1089.2023.19439
中图分类号：
学科分类号：
摘要：
3D generative adversarial networks (3D GANs) are widely utilized in model prediction and object generation. To address the challenges of massive computation and significant energy consumption in accelerating 3D GANs, a novel fast Fourier transform based frequency-domain accelerator (called FAG) is proposed. Firstly, FAG provides a frequency-domain hardware architecture, which utilizes the compact computation complexity and the zero repeat pattern in deconvolution, to reduce the computation overhead in 3D GANs accelerators. Secondly, FAG exploits the frequency-domain characteristic of Hermitian symmetry and the zero repeat pattern of deconvolution to significantly reduce data movements, and it utilizes the repeat pattern in deconvolution to significantly reduce data movements. Comprehensive evaluations based on the ModelNet data-set and three 3D GAN models (3DGAN, 3D-IWGAN and 3D-PhysNet) show that, the performance and the energy efficiency can be improved by 76% and 141%, respectively, compared with the frequency-domain baseline; FAG achieves 6× higher performance and 46× better energy efficiency compared with the spatial baseline. © 2023 Institute of Computing Technology. All rights reserved.
引用
收藏
页码：953 / 960
页数：7
相关论文
共 21 条
  • [1] Mao X D, Li Q, Xie H R, Et al., Least squares generative adversarial networks, Proceedings of the IEEE International Conference on Computer Vision, pp. 2813-2821, (2017)
  • [2] Wu J J, Zhang C K, Xue T F, Et al., Learning a probabilistic latent space of object shapes via 3D generative-adversarial modeling
  • [3] Yazdanbakhsh A, Samadi K, Kim N S, Et al., GANAX: a unified MIMD-SIMD acceleration for generative adversarial networks, Proceedings of the 45th ACM/IEEE Annual International Symposium on Computer Architecture, pp. 650-661, (2018)
  • [4] Good I J., The interaction algorithm and practical Fourier analysis, Journal of the Royal Statistical Society: Series B (Methodological), 20, 2, pp. 361-372, (1958)
  • [5] Tutatchikov V S., Two-dimensional fast Fourier transform: batterfly in analog of Cooley-Tukey algorithm, Proceedings of the 11th International Forum on Strategic Technology, pp. 495-498, (2016)
  • [6] Mathieu M, Henaff M, LeCun Y., Fast training of convolutional networks through FFTs
  • [7] Vasilache N, Johnson J, Mathieu M, Et al., Fast convolutional nets with fbfft: a GPU performance evaluation
  • [8] Pei S C, Ding J J, Chang J H., Efficient implementation of quaternion Fourier transform, convolution, and correlation by 2-D complex FFT, IEEE Transactions on Signal Processing, 49, 11, pp. 2783-2797, (2001)
  • [9] Feng Y, Whatmough P, Zhu Y H., ASV: accelerated stereo vision system, Proceedings of the 52nd Annual IEEE/ACM International Symposium on Microarchitecture, pp. 643-656, (2019)
  • [10] Xu D W, Tu K J, Wang Y, Et al., FCN-engine: accelerating deconvolutional layers in classic CNN processors, Proceedings of the International Conference on Computer-Aided Design, pp. 1-6, (2018)
123