Hausdorff GAN: Improving GAN Generation Quality With Hausdorff Metric

By Wei Li, Zhixuan Liang, Ping Ma, Ruobei Wang, Xiaohui Cui, Ping Chen. 2021

34 citations recorded by the source corpus.

Data usually resides on a manifold, and the minimal dimension of such a manifold is called its <italic>intrinsic dimension</italic>. This fundamental data property is not considered in the generative adversarial network (GAN) model along with its its variants; such that original data and generated data often hold different <italic>intrinsic dimensions</italic>. The different <italic>intrinsic dimensions</italic> of both generated and original data may cause generated data distribution to not match original data distribution completely, and it certainly will hurt the quality of generated data. In this study, we first show that GAN is often unable to generate simulation data, holding the same <italic>intrinsic dimension</italic> as the original data with both theoretical analysis and experimental illustration. Next, we propose a new model, called <italic>Hausdorff</italic> GAN, which removes the issue of different <italic>intrinsic dimensions</italic> and introduces the <italic>Hausdorff</italic> metric into GAN training to generate higher quality data. This provides new insights into the success of <italic>Hausdorff</italic> GAN. Specifically, we utilize a mapping function to map both original and generated data into the same manifold. We then calculate the <italic>Hausdorff</italic> distance to measure the difference between the mapped original data and the mapped generated data, toward pushing generated data to the side of original data. Finally, we conduct extensive experiments (using MNIST, CIFAR10, and CelebA datasets) to demonstrate the significant performance improvement of the <italic>Hausdorff</italic> GAN in achieving the largest <italic>Inception Score</italic> and the smallest Frechet inception distance (FID) score as well as producing diverse generated data at different resolutions.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Generative Models