Abstract

One of the major reasons that deep learning for supervised learning has been so successful is that deep networks allow us to tame high dimensional input spaces and transform them into simpler, lower dimensional objects that we can work with. This talk argues that a similar approach using ML techniques to simplify scenarios to their "low dimensional game theoretic representation" can yield progress in building agents that cooperate and coordinate, as well as allow us to apply mechanism design ideas at large scale.

Video Recording