One interesting aspect of statistical machine learning is the cornucopia of techniques related to the "exploration versus exploitation" dilemma in exploring a g

One interesting aspect of statistical machine learning is the cornucopia of techniques related to the "exploration versus exploitation" dilemma in exploring a giant search space. Cf. "bandit learning" I think it will take a novel combination of many different aspects of data mining, machine learning, and network science to make a major breakthrough, but it also definitely sounds like a really fun problem to work on.