Equitable Multi-task Learning
This addresses the issue of task imbalance in MTL for researchers and practitioners, offering a practical solution with demonstrated gains, though it is incremental as it builds on existing multi-task optimization methods.
The paper tackles the problem of inequitable learning in multi-task learning (MTL) where some tasks are overlooked due to competing correlations, and proposes EMTL, a method that regularizes relative task contributions to improve generalization, achieving stable outperformance over state-of-the-art methods on public benchmarks and significant improvements in multi-task recommendation in A/B tests.
Multi-task learning (MTL) has achieved great success in various research domains, such as CV, NLP and IR etc. Due to the complex and competing task correlation, naive training all tasks may lead to inequitable learning, i.e. some tasks are learned well while others are overlooked. Multi-task optimization (MTO) aims to improve all tasks at same time, but conventional methods often perform poor when tasks with large loss scale or gradient norm magnitude difference. To solve the issue, we in-depth investigate the equity problem for MTL and find that regularizing relative contribution of different tasks (i.e. value of task-specific loss divides its raw gradient norm) in updating shared parameter can improve generalization performance of MTL. Based on our theoretical analysis, we propose a novel multi-task optimization method, named EMTL, to achieve equitable MTL. Specifically, we efficiently add variance regularization to make different tasks' relative contribution closer. Extensive experiments have been conduct to evaluate EMTL, our method stably outperforms state-of-the-art methods on the public benchmark datasets of two different research domains. Furthermore, offline and online A/B test on multi-task recommendation are conducted too. EMTL improves multi-task recommendation significantly, demonstrating the superiority and practicability of our method in industrial landscape.