1: \begin{abstract}

2:   Reducing the amount of human supervision is a key problem in machine learning and a natural  approach is that of exploiting  the relations  (structure) among  different tasks. This is the idea at the core of multi-task learning. In this context a fundamental question is how to incorporate the tasks structure in the learning problem.

3:   We tackle this question by studying a general computational framework that allows to encode a-priori knowledge of the tasks structure in the form of a convex penalty; in this setting a variety of previously proposed methods can be recovered as special cases, including linear and non-linear approaches. Within this framework, we show that tasks and their structure can be efficiently learned considering a convex optimization problem that can be approached by means of block coordinate methods such as alternating minimization and for which we prove convergence to the global minimum.

4: 

5:   

6: %  Our approach allows to consider 

7: %  

8: %  

9: %   {\color{blue}The framework we consider and the meta-optimization method we propose can be naturally expressed in both a dual and primal formulation. This allows for the straightforward extension of previous multi-task methods to the non-linear settings and furthermore provides with practical optimization algorithms to solve them.}

10:   %Another focus of the paper is on methods for learning interpretable tasks structure. To this end we propose a novel setting whose aim is to recover sparse sets of relations among tasks.

11: \end{abstract}

12: