Understanding and Improving Information Transfer in Multi-Task Learning

Abstract

We investigate multi-task learning approaches that use a shared feature representation for all tasks. To better understand the transfer of task information, we study an architecture with a shared module for all tasks and a separate output module for each task. We study the theory of this setting on linear and ReLU-activated models. Our key observation is that whether or not tasks' data are well-aligned can significantly affect the performance of multi-task learning. We show that misalignment between task data can cause negative transfer (or hurt performance) and provide sufficient conditions for positive transfer. Inspired by the theoretical insights, we show that aligning tasks' embedding layers leads to performance gains for multi-task training and transfer learning on the GLUE benchmark and sentiment analysis tasks; for example, we obtain a 2.35% GLUE score average improvement on 5 GLUE tasks over BERT-LARGE using our alignment method. We also design an SVD-based task reweighting scheme and show that it improves the robustness of multi-task training on a multi-label image dataset.

Appeared in ICLR 2020

References in corpus (5)

• An Overview of Multi-Task Learning in Deep Neural Networks
• CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning
• The Natural Language Decathlon: Multitask Learning as Question Answering
• An introduction to domain adaptation and transfer learning
• The Computational Complexity of Training ReLU(s)

Cited by in corpus (15)

• Distribution Matching for Heterogeneous Multi-Task Learning: a Large-scale Face Study
• Efficiently Identifying Task Groupings for Multi-Task Learning
• A Comprehensive Evaluation of Multi-task Learning and Multi-task Pre-training on EHR Time-series Data
• In-N-Out: Pre-Training and Self-Training using Auxiliary Information for Out-of-Distribution Robustness
• On the Complementarity of Data Selection and Fine Tuning for Domain Adaptation
• Scalable Multitask Learning Using Gradient-based Estimation of Task Affinity
• Embedding Adaptation is Still Needed for Few-Shot Learning
• Pay Better Attention to Attention: Head Selection in Multilingual and Multi-Domain Sequence Modeling
• On Localized Discrepancy for Domain Adaptation
• Boosting Multitask Learning on Graphs through Higher-Order Task Affinities
• CrossFit: A Few-shot Learning Challenge for Cross-task Generalization in NLP
• Pareto Navigation Gradient Descent: a First-Order Algorithm for Optimization in Pareto Set
• Comparing Deep Learning strategies for paired but unregistered multimodal segmentation of the liver in T1 and T2-weighted MRI
• Meta-learning Transferable Representations with a Single Target Domain
• Disentangling Transfer and Interference in Multi-Domain Learning