Self-Supervised Generalisation with Meta Auxiliary Learning

Published in

Neural Information Processing Systems(2019)

External Links:

Generate Graph

TL;DR

The proposed method, Meta AuXiliary Learning (MAXL), outperforms single-task learning on 7 image datasets, without requiring any additional data, and is even competitive when compared with human-defined auxiliary labels.

Abstract

Learning with auxiliary tasks can improve the ability of a primary task to generalise. However, this comes at the cost of manually labelling auxiliary data. We propose a new method which automatically learns appropriate labels for an auxiliary task, such that any supervised learning task can be improved without requiring access to any further data. The approach is to train two neural networks: a label-generation network to predict the auxiliary labels, and a multi-task network to train the primary task alongside the auxiliary task. The loss for the label-generation network incorporates the loss of the multi-task network, and so this interaction between the two networks can be seen as a form of meta learning with a double gradient. We show that our proposed method, Meta AuXiliary Learning (MAXL), outperforms single-task learning on 7 image datasets, without requiring any additional data. We also show that MAXL outperforms several other baselines for generating auxiliary labels, and is even competitive when compared with human-defined auxiliary labels. The self-supervised nature of our method leads to a promising new direction towards automated generalisation. Source code can be found at this https URL.

Authors

A. Davison

7 papers

Shikun Liu

2 papers

Edward Johns

1 papers

References40 items

Learning to Generalize from Sparse and Underspecified Rewards

Adapting Auxiliary Losses Using Gradient Similarity

Fine-Grained Visual Categorization using Meta-Learning Optimization with Sample Selection of Auxiliary Data

Auxiliary Tasks in Multi-task Learning

End-To-End Multi-Task Learning With Attention

Self-Supervised Generalisation with Meta Auxiliary Learning

Published in

Neural Information Processing Systems(2019)

External Links:

Generate Graph

TL;DR

Abstract

Authors

A. Davison

7 papers

Shikun Liu

2 papers

Edward Johns

1 papers

References40 items

Learning to Generalize from Sparse and Underspecified Rewards

Adapting Auxiliary Losses Using Gradient Similarity

Fine-Grained Visual Categorization using Meta-Learning Optimization with Sample Selection of Auxiliary Data

Auxiliary Tasks in Multi-task Learning

End-To-End Multi-Task Learning With Attention

Multi-task Self-Supervised Visual Learning

Focal Loss for Dense Object Detection

Meta-SGD: Learning to Learn Quickly for Few Shot Learning

An Overview of Multi-Task Learning in Deep Neural Networks

Unsupervised Learning of Depth and Ego-Motion from Video

Multitask Learning with Low-Level Auxiliary Tasks for Encoder-Decoder Based Speech Recognition

Prototypical Networks for Few-shot Learning

Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks

Reinforcement Learning with Unsupervised Auxiliary Tasks

Optimization as a Model for Few-Shot Learning

UberNet: Training a Universal Convolutional Neural Network for Low-, Mid-, and High-Level Vision Using Diverse Datasets and Limited Memory

Decoupled Neural Interfaces using Synthetic Gradients

Meta-Learning with Memory-Augmented Neural Networks

Learning to learn by gradient descent by gradient descent

Matching Networks for One Shot Learning

One-shot Learning with Memory-Augmented Neural Networks

Cross-Stitch Networks for Multi-task Learning

Deep Residual Learning for Image Recognition

Deep Stereo: Learning to Predict New Views from the World's Imagery

Very Deep Convolutional Networks for Large-Scale Image Recognition

Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks

UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild

An Analysis of Single-Layer Networks in Unsupervised Feature Learning

ImageNet: A large-scale hierarchical image database

Multitask Learning

Learning Complex, Extended Sequences Using the Principle of History Compression

Learning a synaptic learning rule

A k-means clustering algorithm

Learning Multiple Layers of Features from Tiny Images

Visualizing Data using t-SNE

On the Optimization of a Synaptic Learning Rule

The Pascal Visual Object Classes Challenge 2006 ( VOC 2006 ) Results

GradientBased Learning Applied to Document Recognition

Gradient-based learning applied to document recognition

School of Informatics, University of Edinburgh

Field of Study

Computer ScienceMathematics

Venue Information

Name

Neural Information Processing Systems

Type

conference

URL

http://neurips.cc/

Alternate Names

Neural Inf Process Syst
NeurIPS
NIPS