Unsaturated gameIt is inspired by heuristic methods rather than theory-based analysis.
The only motivation for this version of the game is to ensure a stable gradient even when a player faces failure.

In this version of the game, the game is no longer zero-sum and cannot be described using a single value function.

【1】https://sinpycn.github.io/2017/05/03/GAN-Tutorial-How-do-GANs-work.html

【2】https://liqiang311.github.io/deeplearning/generative adversarial network GAN/