Pareto Multi-Task Learning

Xi Lin, Huiling Zoe Zhen, Zhenhua Li, Qing-Fu Zhang, Sam Kwong

Abstract: Multi-task learning is a powerful method for solving multiple correlated tasks simultaneously. However, it is often difficult to combine all the tasks together for finding a single best solution, since different tasks might conflict with each other. Recently, a novel method is proposed to find one solution with good trade-off among different tasks by formulating multi-task learning as multi-objective optimization. In this paper, we generalize this idea and propose a novel Pareto multi-task learning algorithm (Pareto MTL) to find a set of widely distributed Pareto solutions which can represent different trade-offs among different tasks. The proposed algorithm first formulates multi-task learning as a multi-objective optimization problem, and then decomposes it into a set of constrained multi-objective subproblems with different trade-off preferences. By solving these multi-objective subproblems in parallel, Pareto MTL can provide a set of well-representative Pareto solutions with different trade-offs among all tasks. Practitioners can easily select their preferred solution based on all Pareto solutions' performance, or use different trade-off solutions for different situations. Experimental results confirm that the proposed algorithm can generate well-representative solutions and outperform some state-of-the-art algorithms on many multi-task learning applications.

Code Link: https://github.com/Xi-L/ParetoMTL

CMT Num: 6489