Abstract
Network architecture search achieves state-of-the-art results in various tasks such as classification and semantic segmentation. Recently, a reinforcement learning-based approach has been proposed for generative adversarial networks (GANs) search. In this work, we propose an alternative strategy for GAN search by using a proxy task instead of common GAN training. Our method is called differentiable efficient generator search, which focuses on efficiently finding the generator in the GAN. Our search algorithm is inspired by the differential architecture search strategy and the global latent optimization procedure. This leads to both an efficient and stable GAN search. After the generator architecture is found, it can be plugged into any existing framework for GAN training. For consistency-term GAN, which we use in this work, the new model outperforms the original inception score results by 0.25 for CIFAR-10.
Similar content being viewed by others
Availability of data and material
Open-source data were used in this paper
References
Abu Hussein S, Tirer T, Giryes R (2020) Image-adaptive GAN based reconstruction. In: AAAI conference on artificial intelligence
Andrychowicz M, Denil M, Gomez Colmenarejo S, Hoffman MW, Pfau D, Schaul T, Shillingford B, de Freitas N (2016) Learning to learn by gradient descent by gradient descent. In: Advances in neural information processing systems (NIPS)
Arjovsky M, Chintala S, Bottou L (2017) Wasserstein gan. arXiv:1701.07875
Bojanowski P, Joulin A, Lopez Paz D, Szlam A (2018) Optimizing the latent space of generative networks. In: International conference on machine learning (ICML)
Bora A, Jalal A, Price E, Dimakis AG (2017) Compressed sensing using generative models. PMLR, International Convention Centre, Sydney, Australia, pp 537–546
Brock A, Donahue J, Simonyan K (2018) Large scale gan training for high fidelity natural image synthesis. arXiv:1809.11096v2
Cai H, Zhu L, Han S (2019) Proxylessnas: direct neural architecture search on target task and hardware. In: ICLR
Chen X, Xie L, Wu J, Tian Q (2019) Progressive differentiable architecture search: Bridging the depth gap between search and evaluation. arXiv:1904.12760
Chen Y, Hoffman MW, Colmenarejo SG, Denil M, Lillicrap TP, Botvinick M, de Freitas N (2017) Learning to learn without gradient descent by gradient descent. In: International conference on machine learning (ICML)
Clark A, Donahue J, Simonyan K (2019) Adversarial video generation on complex datasets. arXiv:1907.06571
Coates A, Ng A, Lee H (2011) An analysis of single-layer networks in unsupervised feature learning. In: Proceedings of the fourteenth international conference on artificial intelligence and statistics. JMLR workshop and conference proceedings, pp 215–223
Cubuk ED, Zoph B, Mane D, Vasudevan V, Viet Le Q (2018) Autoaugment: Learning augmentation policies from data. arXiv:1805.09501v2
Feigin Y, Spitzer H, Giryes R (2020) Gmm-based generative adversarial encoder learning
Geifman Y, El-Yaniv R (2019) Deep active learning with a neural architecture search. In: NeurIPS
Gong X, Chang S, Jiang Y, Wang Z (2019) Autogan: Neural architecture search for generative adversarial networks. In: ICCV
Goodfellow I, Pouget-Abadie J, Mirza M, Xu B, Warde-Farley D, Ozair S, Courville A, Bengio Y (2014) Generative adversarial nets. In: Advances in neural information processing systems, pp 2672–2680
Gulrajani I, Ahmed F, Arjovsky M, Dumoulin V, Courville A (2017) Improved training of wasserstein gans. arXiv:1704.00028
He K, Zhang X, Ren S, Sun J (2015) Deep residual learning for image recognition. arXiv:1512.03385
Heusel M, Ramsauer H, Unterthiner T, Nessler B, Hochreiter S (2017) Gans trained by a two time-scale update rule converge to a local nash equilibrium. In: Advances in neural information
Hoshen Y, Wolf L (2018) Nam: Non-adversarial unsupervised domain mapping. In: ECCV, pp 455–470
Huang Y, Cheng Y, Chen D, Lee H, Ngiam J, Viet Le Q, Chen, Z (2018) Gpipe: Efficient training of giant neural networks using pipeline parallelism. arXiv:1811.06965
Isola P, Zhu JY, Zhou T, Efros AA (2017) Image-to-image translation with conditional adversarial networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 1125–1134
Kaplan S, Giryes R (2020) Self-supervised neural architecture search
Karras T, Aila T, Laine S, Lehtinen J (2017) Progressive growing of gans for improved quality, stability, and variation. In: ICLR
Karras T, Laine S, Aila T (2019) A style-based generator architecture for generative adversarial networks. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp 4401–4410
Krizhevsky A, Hinton G (2009) Learning multiple layers of features from tiny images. University of Toronto, Toronto
Kurach K, Lucic M, Zhai X, Michalski M, Gelly S (2019) A large-scale study on regularization and normalization in gans. Int Conf Mach Learn 97:3581–3590
Liu C, Chen LC, Schroff F, Adam H, Hua W, Yuille A, Li FF (2019) Auto-deeplab: hierarchical neural architecture search for semantic image segmentation. In: CVPR
Liu C, Dollár P, He K, Girshick R, Yuille A, Xie S (2020) Are labels necessary for neural architecture search? In: ECCV
Liu H, Simonyan K, Yang Y (2019) Darts: differentiable architecture search. In: International conference on learning representations (ICLR)
Liu Z, Luo P, Wang X, Tang X (2018) Large-scale celebfaces attributes (celeba) dataset. Retrieved August 15(2018):11
Miyato T, Kataoka T, Koyama M, Yoshida Y (2018) Spectral normalization for generative adversarial networks. In: ICLR
Ni Y, Song D, Zhang X, Wu H, Liao L (2018) Cagan: Consistent adversarial training enhanced gans. In: International joint conference on artificial intelligence (IJCAI)
Ning W, Yang G, Hao C, Peng W, Zhi T, Chunhua S (2019) Nas-fcos: Fast neural architecture search for object detection. arXiv:1906.04423
Noy A, Nayman N, Ridnik T, Zamir N, Doveh S, Friedman I, Giryes R, Zelnik-Manor L (2019) Asap: architecture search, anneal and prune. arXiv:1904.04123
Odena A, Olah C, Shlens J (2016) Conditional image synthesis with auxiliary classifier gans. In: ICML
Pham H, Guan MY, Zoph B, Viet Le Q, Dean J (2018) Efficient neural architecture search via parameter sharing. In: International conference on machine learning (ICML)
Radford A, Metz L, Chintala S (2016) Unsupervised representation learning with deep convolutional generative adversarial networks. arXiv:1511.06434
Raj A, Li Y, Bresler Y (2019) Gan-based projector for faster recovery with convergence guarantees in linear inverse problems. In: 2019 IEEE/CVF international conference on computer vision (ICCV), pp 5601–5610. https://doi.org/10.1109/ICCV.2019.00570
Real E, Aggarwal A, Huang Y, Viet Le Q (2018) Regularized evolution for image classifier architecture search. In: International conference on machine learning-ICML AutoML workshop
Real E, Moore S, Selle A, Saxena S, Suematsu YL, Tan J, Viet Le Q, Kurakin A (2017) Large-scale evolution of image classifiers. In: International conference on machine learning (ICML)
Rubin M, Stein O, Turko NA, Nygate Y, Roitshtain D, Karako L, Barnea I, Giryes R, Shaked NT (2019) Top-gan: stain-free cancer cell classification using deep learning with a small training set. Med Image Anal 57:176–185
Salimans T, Goodfellow I, Zaremba W, Cheung V, Radford A, Chen X (2016) Improved techniques for training gans. In: Advances in neural information processing systems, pp 2234–2242
Sun G, Ding S, Sun T, Zhang C (2021) SA-CapsGAN: using capsule networks with embedded self-attention for generative adversarial network. Neurocomputing 423:399–406. https://doi.org/10.1016/j.neucom.2020.10.092
Tulyakov S, Liu MY, Yang X, Kautz J (2018) Mocogan: decomposing motion and content for video generation. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 1526–1535
Wang H, Huan J (2019) Agan: towards automated design of generative adversarial networks. arXiv:1906.11080
Wei X, Gong B, Liu Z, Lu W, Wang L (2018) Improving the improved training of wasserstein gans: a consistency term and its dual effect. In: International conference on learning representation (ICLR)
Weng Y, Zhou T, Li Y, Qiu X (2019) Nas-unet: neural architecture search for medical image segmentation. In: IEEE Access, vol 7
Yan S, Zheng Y, Ao W, Zeng X, Zhang M (2020) Does unsupervised architecture representation learning help neural architecture search? NeurIPS 33
Zhang J, Ding S, Zhang N, Jia W (2020) Adversarial training methods for Boltzmann machines. IEEE Access 8:4594–4604. https://doi.org/10.1109/ACCESS.2019.2962758
Zhu JY, Park T, Isola P, Efros AA (2017) Unpaired image-to-image translation using cycle-consistent adversarial networks. In: Proceedings of the IEEE international conference on computer vision, pp 2223–2232
Zhu Z, Liu C, Yang D, Yuille A, Xu D (2019) V-nas: neural architecture search for volumetric medical image segmentation. arXiv:1906.0281
Zoph B, Viet Le Q (2017) Neural architecture search with reinforcement learning. In: International conference on learning representations (ICLR)
Acknowledgements
This work was supported by Alibaba and the NSF-BSF grant.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
IBM-Research intern.
Code availability
Code will be released upon acceptance.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Doveh, S., Giryes, R. DEGAS: differentiable efficient generator search. Neural Comput & Applic 33, 17173–17184 (2021). https://doi.org/10.1007/s00521-021-06309-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00521-021-06309-8