Minimax Multi-Task Learning

A multi-task learning (MTL) algorithm learns an inductive bias to learn several tasks together. MTL is incredibly pervasive in machine learning: it has natural connections to random effects models [5]; user preference prediction (including collaborative filtering) can be framed as MTL [6]; multi-class classification admits the popular one-vs-all and all-pairs MTL reductions; and MTL admits provably good learning in settings where single-task learning is hopeless [3, 4]. But if we see a random set of tasks today, which of the tasks will matter tomorrow? Not knowing the challenges nature will pose in the future, it is wise to mitigate the worst case by ensuring a minimum proficiency on each task.