Beyond Shared Hierarchies: Deep Multitask Learning through Soft Layer Ordering

Abstract

Existing deep multitask learning (MTL) approaches align layers shared between tasks in a parallel ordering. Such an organization significantly constricts the types of shared structure that can be learned. The necessity of parallel ordering for deep MTL is first tested by comparing it with permuted ordering of shared layers. The results indicate that a flexible ordering can enable more effective sharing, thus motivating the development of a soft ordering approach, which learns how shared layers are applied in different ways for different tasks. Deep MTL with soft ordering outperforms parallel ordering methods across a series of domains. These results suggest that the power of deep MTL comes from learning highly general building blocks that can be assembled to meet the demands of each task.

14 pages (main paper: 10 pages). Published as a conference paper at ICLR 2018

References in corpus (2)

• PathNet: Evolution Channels Gradient Descent in Super Neural Networks
• Universal representations:The missing link between faces, text, planktons, and cat breeds

Cited by in corpus (8)

• Multi-Task Learning with Deep Neural Networks: A Survey
• AdaShare: Learning What To Share For Efficient Deep Multi-Task Learning
• Recent Advances of Continual Learning in Computer Vision: An Overview
• Modular meta-learning
• Exploring Shared Structures and Hierarchies for Multiple NLP Tasks
• Feature Partitioning for Efficient Multi-Task Architectures
• Modular meta-learning in abstract graph networks for combinatorial generalization
• Fast Line Search for Multi-Task Learning