To reach the pareto optimal point  ( minimax = maxmin) for the huge ranges of combinatorial optimization problems, I am taking an unorthodox approach that is using GANs (Generative Adversarial Networks).

I am using 2 deep neural networks, one is called Generator Network & another is called Discriminator Networks. The Generator will propose the best possible solution and Discriminator will discriminates the solutions saying that the better solution exists. As epochs goes, it will ultimately converge and we shall get minimax = maxmin solution and pareto optimal will be reached.

This solution will be the best solution for the optimization problem.