CDE-GAN: Cooperative Dual Evolution Based Generative Adversarial Network

Shiming Chen¹, Wenjie Wang¹, Beihao Xia¹, Xinge You¹, Qinmu Peng¹, Zehong Cao², and Weiping Ding³

¹Huazhong University of Science and Technology (HUST), China

²University of Tasmania (UTAS), Australia     ³Nantong University (NTU), China

{shimingchen, wangwj54, xbh_hust, youxg, pengqinmu}@hust.edu.cn   zehong.cao@utas.edu.au   ding.wp@ntu.edu.cn

Abstract

Generative adversarial networks (GANs) have been a popular deep generative model for real-world applications. Despite many recent efforts on GANs that have been contributed, mode collapse and instability of GANs are still open problems caused by their adversarial optimization difficulties. In this paper, motivated by the cooperative co-evolutionary algorithm, we propose a \textit{Cooperative Dual Evolution based Generative Adversarial Network} (CDE-GAN) to circumvent these drawbacks. In essence, CDE-GAN incorporates dual evolution with respect to the generator(s) and discriminators into a unified evolutionary adversarial framework to conduct effective adversarial multi-objective optimization. Thus it exploits the complementary properties and injects dual mutation diversity into training to steadily diversify the estimated density in capturing multi-modes and improve generative performance. Specifically, CDE-GAN decomposes the complex adversarial optimization problem into two subproblems (generation and discrimination), and each subproblem is solved with a separated subpopulation (\textit{E-Generators} and \textit{E-Discriminators}), evolved by its own evolutionary algorithm. Additionally, we further propose a \textit{Soft Mechanism} to balance the trade-off between E-Generators and E-Discriminators to conduct steady training for CDE-GAN. Extensive experiments on one synthetic dataset and three real-world benchmark image datasets demonstrate that the proposed CDE-GAN achieves a competitive and superior performance in generating good quality and diverse samples over baselines. The code and more generated results are available at our project homepage https://shiming-chen.github.io/CDE-GAN-website/CDE-GAN.html.

Model pipeline

Material

Paper

Code

Evaluation

There are three different benchmark datasets used for evaluating our method, i.e., CIFAR-10, LSUN-Bedrooms, and CelebA. Some of these generated images are used in paper, in which supports better vision quality. Indeed, we also evaluate our method on synthesis dataset.

Experiment 1: Experiments on CIFAR-10 for hyper-parameters analysis

Experiments on the CIFAR-10 dataset for hyper-parameters analysis. (a) Inception score evaluation for different CDE-GANs with various balance factor γ={1,0.5,0.1,0.01}. (b) Inception score evaluation for different CDE-GANs with various numbers of discriminators I={1,2,4,8}.

Experiment 2: Generative Performance Evaluation

2.1 Generated results on Synthetic dataset

Dynamic results of Gaussian kernel estimation over generator iteration for different GANs. For each pair of images, the left one is data distribution (real data is represented in blue, generated data is represented in red), and the right one is KDE plots of the generated data corresponded to its left generated data. From top to bottom, the rows are the results of original GAN, NS-GAN, LSGAN, E-GAN, and CDE-GAN (Ours).

(a) Gaussian kernel estimation with MLP of 3 layers

(b) Gaussian kernel estimation with MLP of 4 layers

2.2 Generated results on real-world datasets

(a) CelebA

(b) CIFAR-10

(c) LSUN-Bedrooms

Experiment 3: Comparisons with state-of-the-art methods

3.1 Quantitative comparison on CIFAR-10 dataset

     

3.2 Qualitative comparison on CelebA

Samples generated by different methods on various natural image datasets. The samples generated by different methods are provided by the original literatures, i.e., MAD-GAN [23], Lipizzaner [24], Mustangs [27], Stabilizing-GAN [10], and acGAN [19].

Acknowledgement

This work was supported in part by the National Natural Science Foundation of China~(61571205 and 61772220), the Key Program for International S\&T Cooperation Projects of China~(2016YFE0121200), the Special Projects for Technology Innovation of Hubei Province~(2018ACA135), the Key Science and Technology Innovation Program of Hubei Province~(2017AAA017), the Natural Science Foundation of Hubei Province~(2018CFB691), fund from Science, Technology and Innovation Commission of Shenzhen Municipality~(JCYJ20180305180637611, JCYJ20180305180804836 and JSGG20180507182030600). We thank Dr. Chaoyue Wang for his assistance with coding and theoretical suggestions.

Reference

[1] Chaoyue Wang et al. "Evolutionary generative adversarial networks." IEEE Transactions on Evolutionary Computation (TEVC), 2019.

[2] Isabela Albuquerque et al. "Multi-objective training of Generative Adversarial Networks with multiple discriminators." In ICML, 2019.

[3] Jamal Toutouh et al. "Spatial evolutionary generative adversarial networks." In GECCO, 2019.

[4] Tom Schmiedlechner et al. "Towards distributed coevolutionary gans." In AAAI, 2018.

[5] Ishan Durugkar et al. "Generative multi-adversarial networks." In ICLR, 2017.

[6] Ian J. Goodfellow et al. "Generative adversarial nets." In NIPS, 2014.

Top