Multitask Learning with Single Gradient Step Update for Task Balancing
This addresses the imbalance issue in multitask learning for computer vision applications, representing an incremental improvement.
The paper tackles the task imbalance problem in multitask learning by proposing a gradient-based meta-learning algorithm that trains shared and task-specific layers separately, achieving state-of-the-art performance on various computer vision problems.
Multitask learning is a methodology to boost generalization performance and also reduce computational intensity and memory usage. However, learning multiple tasks simultaneously can be more difficult than learning a single task because it can cause imbalance among tasks. To address the imbalance problem, we propose an algorithm to balance between tasks at the gradient level by applying gradient-based meta-learning to multitask learning. The proposed method trains shared layers and task-specific layers separately so that the two layers with different roles in a multitask network can be fitted to their own purposes. In particular, the shared layer that contains informative knowledge shared among tasks is trained by employing single gradient step update and inner/outer loop training to mitigate the imbalance problem at the gradient level. We apply the proposed method to various multitask computer vision problems and achieve state-of-the-art performance.