Evolving GAN Formulations for Higher Quality Image Synthesis (2023)
Generative Adversarial Networks (GANs) have extended deep learning to complex generation and translation tasks across different data modalities. However, GANs are notoriously difficult to train: Mode collapse and other instabilities in the training process often degrade the quality of the generated results, such as images. This paper presents a new technique called TaylorGAN for improving GANs by discovering customized loss functions for each of its two networks. The loss functions are parameterized as Taylor expansions and optimized through multiobjective evolution. On an image-to-image translation benchmark task, this approach qualitatively improves generated image quality and quantitatively improves two independent GAN performance metrics. It therefore forms a promising approach for applying GANs to more challenging tasks in the future.
To Appear In R. Kozma, C. Alippi, Y. Choe, and F. C. Morabito, editors, Artificial Intelligence in the Age of Neural Networks and Brain Computing (second edition), New York, 2023. Elsevier. Also arXiv:2102.08578.

Santiago Gonzalez Ph.D. Alumni slgonzalez [at] utexas edu
Risto Miikkulainen Faculty risto [at] cs utexas edu