Theoretical properties of generative adversarial networks (GANs)

This project explored the theoretical properties of generative adversarial networks, regarding their convergence.

• We analyzed the conditions under which GANs converge to an equilibrium, ensuring that the generator learns to produce realistic data distributions.
• We studied a specific cases of mode collapse, where the generator fails to capture the full diversity of the data, and showed the mathematical theorems.

Toolkit for the project

• R, R Studio, LaTeX.