You can use GANs to generate fake data based on stuff like particle accelerator data or electronic health records. Whether you can use StyleGAN specifically is unclear. What's the equivalent of progressive growing on tabular/numeric data? Or style transfer?