conftrace_

2020 ICML ICML 2020

Adaptive Adversarial Multi-task Representation Learning

Abstract

Adversarial Multi-task Representation Learning (AMTRL) methods are able to boost the performance of Multi-task Representation Learning (MTRL) models. However, the theoretical mechanism behind AMTRL is less investigated. To fill this gap, we study the generalization error bound of AMTRL through the lens of Lagrangian duality . Based on the duality, we proposed an novel adaptive AMTRL algorithm which improves the performance of original AMTRL methods. The extensive experiments back up our theoretical analysis and validate the superiority of our proposed algorithm.

Authors

Yuren Mao, Weiwei Liu, Xuemin Lin

Topics

Artificial Intelligence > Learning Paradigms > Transfer Learning Machine Learning > Core Methods > Representation Learning Machine Learning > Learning Types > Adversarial Learning Machine Learning > Learning Types > Multi-Task Learning Machine Learning > Learning Types > Transfer Learning Deep Learning > Learning Types > Multimodal Learning

Keywords

representation learning adversarial learning multi-task learning transfer learning gradient-based optimization generalization error bound generalization bound multi-task representation learning adversarial multi-task lagrangian duality reward transformation

More from ICML 2020

All ICML 2020 papers →

From Local SGD to Local Fixed-Point Methods for Federated Learning

Adaptive Gradient Descent without Descent

Emergence of Separable Manifolds in Deep Language Representations

On Learning Sets of Symmetric Elements

Stochastically Dominant Distributional Reinforcement Learning

Minimax Pareto Fairness: A Multi Objective Perspective