Abstract
Despite the growing prominence of generative adversarial networks (GANs), improving the performance of GANs is still a challenging problem. To this end, a combination method for training GANs is proposed by coupling spectral normalization with a zero-centered gradient penalty technique (the penalty is done on the inner function of Sigmoid function of discriminator). Particularly, the proposed method not only overcomes the limitations of networks convergence and training instability but also alleviates the mode collapse behavior in GANs. Experimentally, the improved method becomes more competitive compared with some of recent methods on several datasets.
Sopported by the National Natural Science Foundation of China: Managing Uncertainty in Service Based Software with Intelligent Adaptive Architecture (No. 61732019).
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Arjovsky, M., Bottou, L.: Towards principled methods for training generative adversarial networks. arxiv e-prints, art. arXiv preprint arXiv:1701.04862 (2017)
Arjovsky, M., Chintala, S., Bottou, L.: Wasserstein gan. arXiv preprint arXiv:1701.07875 (2017)
Bansal, N., Chen, X., Wang, Z.: Can we gain more from orthogonality regularizations in training deep CNNS? In: Proceedings of the 32nd International Conference on Neural Information Processing Systems, pp. 4266–4276. Curran Associates Inc. (2018)
Barratt, S., Sharma, R.: A note on the inception score. arXiv preprint arXiv:1801.01973 (2018)
Brock, A., Donahue, J., Simonyan, K.: Large scale gan training for high fidelity natural image synthesis. arXiv preprint arXiv:1809.11096 (2018)
Coates, A., Ng, A., Lee, H.: An analysis of single-layer networks in unsupervised feature learning. In: Proceedings of the Fourteenth International Conference on Artificial Intelligence and Statistics, pp. 215–223 (2011)
Daskalakis, C., Ilyas, A., Syrgkanis, V., Zeng, H.: Training gans with optimism. arXiv preprint arXiv:1711.00141 (2017)
Fedus, W., Rosca, M., Lakshminarayanan, B., Dai, A.M., Mohamed, S., Goodfellow, I.: Many paths to equilibrium: Gans do not need to decrease a divergence at every step. arXiv preprint arXiv:1710.08446 (2017)
Glorot, X., Bengio, Y.: Understanding the difficulty of training deep feedforward neural networks. In: Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, pp. 249–256 (2010)
Goodfellow, I., et al.: Generative adversarial nets. In: Advances in Neural Information Processing Systems, pp. 2672–2680 (2014)
Gulrajani, I., Ahmed, F., Arjovsky, M., Dumoulin, V., Courville, A.C.: Improved training of wasserstein gans. In: Advances in Neural Information Processing Systems, pp. 5767–5777 (2017)
van Handel, R.: Probability in high dimension. Technical report (2014)
Heusel, M., Ramsauer, H., Unterthiner, T., Nessler, B., Hochreiter, S.: Gans trained by a two time-scale update rule converge to a local nash equilibrium. In: Advances in Neural Information Processing Systems, pp. 6626–6637 (2017)
Ioffe, S., Szegedy, C.: Batch normalization: Accelerating deep network training by reducing internal covariate shift. arXiv preprint arXiv:1502.03167 (2015)
Isola, P., Zhu, J.Y., Zhou, T., Efros, A.A.: Image-to-image translation with conditional adversarial networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1125–1134 (2017)
Karras, T., Aila, T., Laine, S., Lehtinen, J.: Progressive growing of gans for improved quality, stability, and variation. arXiv preprint arXiv:1710.10196 (2017)
Kingma, D.P., Ba, J.: Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014)
Krizhevsky, A., Hinton, G., et al.: Learning multiple layers of features from tiny images (2009)
Kurach, K., Lucic, M., Zhai, X., Michalski, M., Gelly, S.: A large-scale study on regularization and normalization in gans. arXiv preprint arXiv:1807.04720 (2018)
Liu, Z., Luo, P., Wang, X., Tang, X.: Deep learning face attributes in the wild. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 3730–3738 (2015)
Mao, X., Li, Q., Xie, H., Lau, R.Y., Wang, Z., Paul Smolley, S.: Least squares generative adversarial networks. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 2794–2802 (2017)
Mescheder, L., Geiger, A., Nowozin, S.: Which training methods for gans do actually converge? arXiv preprint arXiv:1801.04406 (2018)
Metz, L., Poole, B., Pfau, D., Sohl-Dickstein, J.: Unrolled generative adversarial networks. arXiv preprint arXiv:1611.02163 (2016)
Miyato, T., Kataoka, T., Koyama, M., Yoshida, Y.: Spectral normalization for generative adversarial networks. arXiv preprint arXiv:1802.05957 (2018)
Nagarajan, V., Kolter, J.Z.: Gradient descent gan optimization is locally stable. In: Advances in Neural Information Processing Systems, pp. 5585–5595 (2017)
Netzer, Y., Wang, T., Coates, A., Bissacco, A., Wu, B., Ng, A.Y.: Reading digits in natural images with unsupervised feature learning (2011)
Oberman, A.M., Calder, J.: Lipschitz regularized deep neural networks converge and generalize. arXiv preprint arXiv:1808.09540 (2018)
Paszke, A., et al.: Automatic differentiation in pytorch (2017)
Petzka, H., Fischer, A., Lukovnicov, D.: On the regularization of wasserstein gans. arXiv preprint arXiv:1709.08894 (2017)
Radford, A., Metz, L., Chintala, S.: Unsupervised representation learning with deep convolutional generative adversarial networks. arXiv preprint arXiv:1511.06434 (2015)
Salimans, T., Goodfellow, I., Zaremba, W., Cheung, V., Radford, A., Chen, X.: Improved techniques for training gans. In: Advances in Neural Information Processing Systems, pp. 2234–2242 (2016)
Saxe, A.M., McClelland, J.L., Ganguli, S.: Exact solutions to the nonlinear dynamics of learning in deep linear neural networks. arXiv preprint arXiv:1312.6120 (2013)
Scardapane, S., Comminiello, D., Hussain, A., Uncini, A.: Group sparse regularization for deep neural networks. Neurocomputing 241, 81–89 (2017)
Thanh-Tung, H., Tran, T., Venkatesh, S.: Improving generalization and stability of generative adversarial networks. arXiv preprint arXiv:1902.03984 (2019)
Tzeng, E., Hoffman, J., Saenko, K., Darrell, T.: Adversarial discriminative domain adaptation. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 7167–7176 (2017)
Virmaux, A., Scaman, K.: Lipschitz regularity of deep neural networks: analysis and efficient estimation. In: Advances in Neural Information Processing Systems, pp. 3835–3844 (2018)
Yeh, R., Chen, C., Lim, T.Y., Hasegawa-Johnson, M., Do, M.N.: Semantic image inpainting with perceptual and contextual losses. arXiv preprint arXiv:1607.07539 2(3) (2016)
Yu, F., Seff, A., Zhang, Y., Song, S., Funkhouser, T., Xiao, J.: Lsun: Construction of a large-scale image dataset using deep learning with humans in the loop. arXiv preprint arXiv:1506.03365 (2015)
Zhang, M., Lucas, J., Ba, J., Hinton, G.E.: Lookahead optimizer: k steps forward, 1 step back. In: Advances in Neural Information Processing Systems, pp. 9593–9604 (2019)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendices
A Training Details on Synthetic Datasets
The 8 Gaussians dataset is sampled from a mixture of 8 Gaussians of standard deviation 0.02, this means are equally spaced around a circle of radius 2. 25 Gaussians dataset, like the 8 Gaussians, is sample from a mixture of 25 Gaussians, which is arranged in a square. Two datasets consist of 100Â k samples. The discriminator contains three SNLinear layers (bias: True, False and True) with 128 hidden units and LReLU (0.2) activation, and the generator contains three Linear layers (bias: False, False and True) with 256 hidden units, BN and ReLU activation.
As for the hyper parameters setting, both networks are optimized using OAdam with a learning rate of 0.0002 and \(\beta _1\,=\,0.5\), \(\beta _2=0.9\) (training the original GAN use Adam). The latent variable \(\textit{\textbf{z}}\sim N(\textit{\textbf{0}}, \textit{\textbf{I}}_{128})\) and the penalty coefficient \(\lambda =10\) with Lipschitz constant \(L=0\). The batchsize is set to 100.
B Networks Architecture on Benchmark Datasets
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this paper
Cite this paper
Tan, H., Zhou, L., Wang, G., Zhang, Z. (2020). Improved Performance of GANs via Integrating Gradient Penalty with Spectral Normalization. In: Li, G., Shen, H., Yuan, Y., Wang, X., Liu, H., Zhao, X. (eds) Knowledge Science, Engineering and Management. KSEM 2020. Lecture Notes in Computer Science(), vol 12275. Springer, Cham. https://doi.org/10.1007/978-3-030-55393-7_37
Download citation
DOI: https://doi.org/10.1007/978-3-030-55393-7_37
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-55392-0
Online ISBN: 978-3-030-55393-7
eBook Packages: Computer ScienceComputer Science (R0)