Putting Bayes to Sleep

In online multitask learning a single algorithm faces a collection of interleaved tasks. By addressing these tasks jointly rather than in isolation, the algorithm can discover and exploit similaritie