The paper presents a method for learning sparse representations that are shared across multiple tasks, generalizing the well-known 1-norm regularization used in single-task learning. The method introduces a novel non-convex regularizer to control the number of common features across tasks. The authors prove that the non-convex problem is equivalent to a convex optimization problem, which can be solved using an iterative algorithm that alternates between learning task-specific functions and learning common sparse representations. The algorithm converges to an optimal solution. Additionally, the method can be extended to learn nonlinear representations using kernels. Experiments on simulated and real datasets demonstrate that the proposed method improves performance relative to learning each task independently and identifies a small number of common features across related tasks. The paper also discusses the relationship between the proposed method and other multi-task learning methods.The paper presents a method for learning sparse representations that are shared across multiple tasks, generalizing the well-known 1-norm regularization used in single-task learning. The method introduces a novel non-convex regularizer to control the number of common features across tasks. The authors prove that the non-convex problem is equivalent to a convex optimization problem, which can be solved using an iterative algorithm that alternates between learning task-specific functions and learning common sparse representations. The algorithm converges to an optimal solution. Additionally, the method can be extended to learn nonlinear representations using kernels. Experiments on simulated and real datasets demonstrate that the proposed method improves performance relative to learning each task independently and identifies a small number of common features across related tasks. The paper also discusses the relationship between the proposed method and other multi-task learning methods.