Pareto Multi-Task Learning

By Xi Lin, Hui-Ling Zhen, Zhenhua Li, Qingfu Zhang, Sam Kwong. 2019

440 citations recorded by the source corpus.

Multi-task learning is a powerful method for solving multiple correlated tasks simultaneously. However, it is often impossible to find one single solution to optimize all the tasks, since different tasks might conflict with each other. Recently, a novel method is proposed to find one single Pareto optimal solution with good trade-off among different tasks by casting multi-task learning as multiobjective optimization. In this paper, we generalize this idea and propose a novel Pareto multi-task learning algorithm (Pareto MTL) to find a set of well-distributed Pareto solutions which can represent different trade-offs among different tasks. The proposed algorithm first formulates a multi-task learning problem as a multiobjective optimization problem, and then decomposes the multiobjective optimization problem into a set of constrained subproblems with different trade-off preferences. By solving these subproblems in parallel, Pareto MTL can find a set of well-representative Pareto optimal solutions with different trade-off among all tasks. Practitioners can easily select their preferred solution from these Pareto solutions, 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.

Subfields: Artificial Intelligence, Computational Theory and Mathematics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computational Theory and Mathematics

Citations in the corpus

Same State, Different Task: Continual Reinforcement Learning without Interference (2021, background): “2019) and has been studied in multi-task (for example (Bishop and Svensén 2012; Lin et al. 2019)) and metalearning (for example, (Rajendran, Irpan, and Jang 2020)).”
Multi-task Learning (2020, methodology): “Expanded Analysis of MultiFashion 1-”
Small Towers Make Big Differences (2020, methodology): “Similar to Sener and Koltun (2018) and Lin et al. (2019), we construct MultiMNIST and MultiFashion dataset by extending MNIST (LeCun et al., 1998) and FashionMNIST (Xiao et al., 2017) to multi-task setups.”
Multi-Task Learning with Deep Neural Networks: A Survey (2020, methodology): “Pareto Multi-Task Learning (Lin et al., 2019) takes this extension one step further by generalizing the algorithm proposed in (Sener and Koltun, 2018) in order to compute multiple Pareto optimal solutions.”
Controlled Text Generation as Continuous Optimization with Multiple Constraints (2021, background): “Since we define all objectives using neural networks, the Pareto front in our case is non-convex, where linear combinations of objectives are shown to be unsuccessful in finding good solutions [36, 37, 10] (see figure 1 for an example).”
Multi-Domain Multi-Task Rehearsal for Lifelong Learning (2020, methodology): “Multi-domain learning (Nam and Han 2016; Tang and Jia 2020) and multi-task learning (Lin et al. 2019; Sener and Koltun 2018) have been studied for years.”
A Survey on Negative Transfer (2020, neutral): “Various techniques have been explored to remedy negative interference, such as altering the gradients directly [97], [98], weighting tasks [99], learning task relatedness [100], [101], routing networks [102], [103], and searching for Pareto solutions [104], [105], etc.”
Multi-Task Learning for Dense Prediction Tasks: A Survey (2020, background): “[22], [23] Other (Sec.”
Contextualizing Multiple Tasks via Learning to Decompose (2021, background): “In standard multi-task learning, the correspondence between an example with one of the multiple tasks is explicit [24, 42, 8].”
Efficiently Identifying Task Groupings for Multi-Task Learning (2021, methodology): “In particular, dynamic loss reweighing has achieved performance superior to using fixed loss weights found with extensive hyperparameter search [28, 18, 37, 10, 45, 33].”
Conflict-Averse Gradient Descent for Multi-task Learning (2021, methodology): “Next, to test CAGrad on more complicated neural models, we perform the same set of experiments on the Multi-Fashion+MNIST benchmark [19] with a shrinked LeNet architecture [18] (in which each layer has a reduced number of neurons compared to the original LeNet).”
Heterogeneous Few-Shot Model Rectification With Semantic Mapping (2020, background): “Thus how to balance different tasks [43], [44] is essential.”
A Survey on the Fairness of Recommender Systems (2022, background): “The tradeoff between multiple fairness and recommendation accuracy might be improved by drawing on Pareto optimization [57] and “seesaw phenomenon”–related work such as Reference [89].”
Blind Image Quality Assessment via Vision-Language Correspondence: A Multitask Learning Perspective (2023, methodology): “[28] cast multitask learning as multi-objective optimization, and solved for the optimal loss weighting according to the Karush-Kuhn-Tucher conditions.”
Mitigating Catastrophic Forgetting in Online Continual Learning by Modeling Previous Task Interrelations via Pareto Optimization (2024, background): “As discussed in (Lin et al., 2019; Momma et al., 2022), seemingly unrelated tasks can exhibit significant dependencies.”
Agree to Disagree: Adaptive Ensemble Knowledge Distillation in Gradient Space (2020, methodology): “Following [26, 16], we formulate the learning of the target descent direction as the following optimization problem: where • θ ( τ ) is the parameter of the student network at iteration τ , • ( θ ( τ ) ) is the KD loss regarding the student and the m -th teacher, • d is the optimizing direction…”
Gradient Vaccine: Investigating and Improving Multi-task Optimization in Massively Multilingual Models (2020, background): “…learning task relatedness (Zhang & Yeung, 2012), routing task-specifc paths (Rusu et al., 2016; Rosenbaum et al., 2019), altering gradients directly (Kendall et al., 2018; Chen et al., 2018a; Du et al., 2018; Yu et al., 2020), or searching pareto solutions (Sener & Koltun, 2018; Lin et al., 2019).”