Multi-Task Learning with Deep Neural Networks: A Survey (2020-09-10T00:00:00.000000Z)

TL;DR

An overview of multi-task learning methods for deep neural networks is given, with the aim of summarizing both the well-established and most recent directions within the field.

Abstract

Multi-task learning (MTL) is a subfield of machine learning in which multiple tasks are simultaneously learned by a shared model. Such approaches offer advantages like improved data efficiency, reduced overfitting through shared representations, and fast learning by leveraging auxiliary information. However, the simultaneous learning of multiple tasks presents new design and optimization challenges, and choosing which tasks should be learned jointly is in itself a non-trivial problem. In this survey, we give an overview of multi-task learning methods for deep neural networks, with the aim of summarizing both the well-established and most recent directions within the field. Our discussion is structured according to a partition of the existing deep MTL techniques into three groups: architectures, optimization methods, and task relationship learning. We also provide a summary of common multi-task benchmarks.

Authors

M. Crawshaw

1 papers

TL;DR

Abstract

Authors

References180 items

Multi-Task Reinforcement Learning as a Hidden-Parameter Block MDP

Maximum Roaming Multi-Task Learning

Learning Functions to Study the Benefit of Multitask Learning

Understanding and Improving Information Transfer in Multi-Task Learning

Sharing Knowledge in Multi-Task Deep Reinforcement Learning

Meta-Learning in Neural Networks: A Survey

MTL-NAS: Task-Agnostic Neural Architecture Search Towards General-Purpose Multi-Task Learning

Multi-Task Reinforcement Learning with Soft Modularization

5分で分かる!? 有名論文ナナメ読み：Jacob Devlin et al. : BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding

MTI-Net: Multi-Scale Task Interaction Networks for Multi-Task Learning

Gradient Surgery for Multi-Task Learning

Pareto Multi-Task Learning

12-in-1: Multi-Task Vision and Language Representation Learning

AdaShare: Learning What To Share For Efficient Deep Multi-Task Learning

Learning Sparse Sharing Architectures for Multiple Tasks

A Comprehensive Survey on Transfer Learning

Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning

Solving Rubik's Cube with a Robot Hand

Generalization in multitask deep neural classifiers: a statistical physics approach

Deep Model Transferability from Attribution Maps

A Comparison of Loss Weighting Strategies for Multi task Learning in Deep Neural Networks

Adaptive Scheduling for Multi-Task Learning

Deep Elastic Networks With Model Selection for Multi-Task Learning

Stochastic Filter Groups for Multi-Task CNNs: Learning Specialist and Generalist Convolution Kernels

Feature Partitioning for Efficient Multi-Task Architectures

OmniNet: A unified architecture for multi-modal multi-task learning

Loss-Balanced Task Weighting to Reduce Negative Transfer in Multi-Task Learning

BAM! Born-Again Multi-Task Networks for Natural Language Understanding

Pattern-Affinitive Propagation Across Depth, Surface Normal and Semantic Segmentation

Recursive Routing Networks: Learning to Compose Modules for Language Understanding

Which Tasks Should Be Learned Together in Multi-task Learning?

Meta Reinforcement Learning with Task Embedding and Shared Policy

Multi-task Learning for Multi-modal Emotion Recognition and Sentiment Analysis

Routing Networks and the Challenges of Modular and Compositional Computation

Representation Similarity Analysis for Efficient Task Taxonomy & Transfer Learning

Improving Multi-Task Deep Neural Networks via Knowledge Distillation for Natural Language Understanding

Attentive Single-Tasking of Multiple Tasks

MultiNet++: Multi-Stream Feature Aggregation and Geometric Loss Strategy for Multi-Task Learning

Branched Multi-Task Networks: Deciding what layers to share

Learning Task Relatedness in Multi-Task Learning for Images in Context

Many Task Learning With Task Routing

A Principled Approach for Learning Task Similarity in Multitask Learning

GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering

Task2Vec: Task Embedding for Meta-Learning

Multi-Task Deep Neural Networks for Natural Language Understanding

Multi-Task Learning of Hierarchical Vision-Language Representation

A Hierarchical Multi-task Approach for Learning Embeddings from Semantic Tasks

Modular Networks: Learning to Decompose Neural Computation

Pyramidal Person Re-IDentification via Multi-Loss Dynamic Training

Multi-Task Learning as Multi-Objective Optimization

Task-Embedded Control Networks for Few-Shot Imitation Learning

Efficient Lifelong Learning with A-GEM

Diversity and Depth in Per-Example Routing Models

Adapting Auxiliary Losses Using Gradient Similarity

Multi-task Learning with Sample Re-weighting for Machine Reading Comprehension

Multi-task Deep Reinforcement Learning with PopArt

Dynamic Task Prioritization for Multitask Learning

Exploring Shared Structures and Hierarchies for Multiple NLP Tasks

Modeling Task Relationships in Multi-task Learning with Multi-gate Mixture-of-Experts

A Modulation Module for Multi-task Learning with Applications in Image Retrieval

Automatically Composing Representation Transformations as a Means for Generalization

Modular meta-learning

DARTS: Differentiable Architecture Search

Gradient Adversarial Training of Neural Networks

The Natural Language Decathlon: Multitask Learning as Question Answering

PAD-Net: Multi-tasks Guided Prediction-and-Distillation Network for Simultaneous Depth Estimation and Scene Parsing

SciTaiL: A Textual Entailment Dataset from Science Question Answering

Taskonomy: Disentangling Task Transfer Learning

GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding

Improved Fusion of Visual and Language Representations by Dense Symmetric Co-attention for Visual Question Answering

End-To-End Multi-Task Learning With Attention

Efficient Parametrization of Multi-domain Deep Neural Networks

Learning Synergies Between Pushing and Grasping with Self-Supervised Deep Reinforcement Learning

Evolutionary architecture search for deep multitask networks

The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks

The Lottery Ticket Hypothesis: Training Pruned Neural Networks