It would be interesting to incorporate a GAN for density estimation of heavy‑tailed distributions. This would involve implementing a version of [Pareto GAN: Extending the Representational Power of GANs to Heavy‑Tailed Distributions](https://arxiv.org/abs/2101.09113).