Approximate exploitability: Learning a best response in large games

Open in new window