Cross-stitch Networks for Multi-task Learning
It addresses the need for generalizable multi-task learning methods in recognition tasks, offering a novel approach beyond task-specific architectures.
The paper tackles the problem of multi-task learning in Convolutional Networks by proposing cross-stitch units to learn optimal shared and task-specific representations, resulting in dramatically improved performance over baselines for categories with few training examples.
Multi-task learning in Convolutional Networks has displayed remarkable success in the field of recognition. This success can be largely attributed to learning shared representations from multiple supervisory tasks. However, existing multi-task approaches rely on enumerating multiple network architectures specific to the tasks at hand, that do not generalize. In this paper, we propose a principled approach to learn shared representations in ConvNets using multi-task learning. Specifically, we propose a new sharing unit: "cross-stitch" unit. These units combine the activations from multiple networks and can be trained end-to-end. A network with cross-stitch units can learn an optimal combination of shared and task-specific representations. Our proposed method generalizes across multiple tasks and shows dramatically improved performance over baseline methods for categories with few training examples.