The sharp, the flat and the shallow: Can weakly interacting agents learn to escape bad minima?

Open in new window