1. What is GAN?

GAN stands for Generative Adversarial Nets. It's a framework for estimating generative model, contains two models: a generative model that captures training data distribution to generate new data, and a discriminative model that estimates the probability that a sample came from the train data rather than generative model. They are both multilayer perceptrons.

2. What is a Differentiable Function?

Differentiable meas the derivative exists. If you want to know more details about it, you can click this link: https://www.mathsisfun.com/calculus/differentiable.html

3. How does GAN work?

To be Continued...