HydaLearn: Highly Dynamic Task Weighting for Multi-task Learning with Auxiliary Tasks
This addresses the challenge of optimizing multi-task learning for practitioners by dynamically adjusting task weights to improve model performance, though it appears incremental as it builds on existing MTL methods.
The paper tackled the problem of poor performance in multi-task learning due to constant loss weights by introducing HydaLearn, a dynamic weighting algorithm that adjusts task weights at the mini-batch level, resulting in performance increases on synthetic data and two supervised learning domains.
Multi-task learning (MTL) can improve performance on a task by sharing representations with one or more related auxiliary-tasks. Usually, MTL-networks are trained on a composite loss function formed by a constant weighted combination of the separate task losses. In practice, constant loss weights lead to poor results for two reasons: (i) the relevance of the auxiliary tasks can gradually drift throughout the learning process; (ii) for mini-batch based optimisation, the optimal task weights vary significantly from one update to the next depending on mini-batch sample composition. We introduce HydaLearn, an intelligent weighting algorithm that connects main-task gain to the individual task gradients, in order to inform dynamic loss weighting at the mini-batch level, addressing i and ii. Using HydaLearn, we report performance increases on synthetic data, as well as on two supervised learning domains.