[Planning] Explore GAN (Generative Adversarial Networks) implementation in ML.NET + TensorFlow

For future features planning it'd be worth to explore possibilities in ML.NET enabling GAN (Generative Adversarial Networks) implementation in ML.NET + TensorFlow.

Most popular use cases are related to images generation, such as this cool prototype app:

https://github.com/microsoft/GenStudio
https://gen.studio/

However, GAN can also be applied to data and very interesting cases are related to **synthetic data generation** from a generative adversarial network (GAN) designed to approximate any original training data distribution. For instance, you have a Time Series dataset of just a couple of years and want to generate a larger synthetic dataset, or any other synthetic data generation.

Many more use cases:
https://github.com/nashory/gans-awesome-applications

Question for anyone in the community reading this issue, can you answer this issue with your thoughts about GAN (Generative Adversarial Networks), your own business scenarios and possibilities you might see in ML.NET?