1: \begin{abstract}%   <- trailing '%' for backward compatibility of .sty file

2: Multi-task sparse feature learning aims to improve the generalization performance by exploiting

3: the shared features among tasks.

4: It has been successfully applied to many applications including computer vision and biomedical informatics.

5: Most of the existing multi-task sparse feature learning

6: algorithms are formulated as a convex sparse regularization problem, which is usually suboptimal, due

7: to its looseness for approximating an $\ell_0$-type regularizer.

8: In this paper, we propose a non-convex formulation

9: for multi-task sparse feature learning based on a novel non-convex regularizer.

10: To solve the non-convex optimization problem, we propose a

11: Multi-Stage Multi-Task Feature Learning (MSMTFL) algorithm; we also provide intuitive interpretations,

12: detailed convergence and reproducibility analysis for the proposed algorithm. Moreover, we present a detailed

13: theoretical analysis showing that MSMTFL achieves a better parameter estimation error bound

14: than the convex formulation.

15: Empirical studies on both synthetic and real-world data sets demonstrate the effectiveness of MSMTFL in comparison with

16: the state of the art multi-task sparse feature learning algorithms.

17: \end{abstract}

18: