A Probabilistic Model for Dirty Multi-task Feature Selection

Multi-task feature selection methods often make the hypothesis that learning tasks share relevant and irrelevant features. However, this hypothesis may be too restrictive in practice. For example, the