Few-shot learning with graph neural networks

Victor Garcia, Joan Bruna Estrach

Research output: Contribution to conferencePaper

Abstract

We propose to study the problem of few-shot learning with the prism of inference on a partially observed graphical model, constructed from a collection of input images whose label can be either observed or not. By assimilating generic message-passing inference algorithms with their neural-network counterparts, we define a graph neural network architecture that generalizes several of the recently proposed few-shot learning models. Besides providing improved numerical performance, our framework is easily extended to variants of few-shot learning, such as semi-supervised or active learning, demonstrating the ability of graph-based models to operate well on ‘relational’ tasks.

Original languageEnglish (US)
StatePublished - Jan 1 2018
Event6th International Conference on Learning Representations, ICLR 2018 - Vancouver, Canada
Duration: Apr 30 2018May 3 2018

Conference

Conference6th International Conference on Learning Representations, ICLR 2018
CountryCanada
CityVancouver
Period4/30/185/3/18

Fingerprint

neural network
Neural networks
learning
Supervised learning
Message passing
Prisms
Network architecture
Labels
Neural Networks
Graph
Inference
ability
performance
Learning Model
Active Learning
Problem-Based Learning

ASJC Scopus subject areas

  • Language and Linguistics
  • Education
  • Computer Science Applications
  • Linguistics and Language

Cite this

Garcia, V., & Bruna Estrach, J. (2018). Few-shot learning with graph neural networks. Paper presented at 6th International Conference on Learning Representations, ICLR 2018, Vancouver, Canada.

Few-shot learning with graph neural networks. / Garcia, Victor; Bruna Estrach, Joan.

2018. Paper presented at 6th International Conference on Learning Representations, ICLR 2018, Vancouver, Canada.

Research output: Contribution to conferencePaper

Garcia, V & Bruna Estrach, J 2018, 'Few-shot learning with graph neural networks' Paper presented at 6th International Conference on Learning Representations, ICLR 2018, Vancouver, Canada, 4/30/18 - 5/3/18, .
Garcia V, Bruna Estrach J. Few-shot learning with graph neural networks. 2018. Paper presented at 6th International Conference on Learning Representations, ICLR 2018, Vancouver, Canada.
Garcia, Victor ; Bruna Estrach, Joan. / Few-shot learning with graph neural networks. Paper presented at 6th International Conference on Learning Representations, ICLR 2018, Vancouver, Canada.
@conference{fa28ba8dfae4452e9076f2d87dcc6a08,
title = "Few-shot learning with graph neural networks",
abstract = "We propose to study the problem of few-shot learning with the prism of inference on a partially observed graphical model, constructed from a collection of input images whose label can be either observed or not. By assimilating generic message-passing inference algorithms with their neural-network counterparts, we define a graph neural network architecture that generalizes several of the recently proposed few-shot learning models. Besides providing improved numerical performance, our framework is easily extended to variants of few-shot learning, such as semi-supervised or active learning, demonstrating the ability of graph-based models to operate well on ‘relational’ tasks.",
author = "Victor Garcia and {Bruna Estrach}, Joan",
year = "2018",
month = "1",
day = "1",
language = "English (US)",
note = "6th International Conference on Learning Representations, ICLR 2018 ; Conference date: 30-04-2018 Through 03-05-2018",

}

TY - CONF

T1 - Few-shot learning with graph neural networks

AU - Garcia, Victor

AU - Bruna Estrach, Joan

PY - 2018/1/1

Y1 - 2018/1/1

N2 - We propose to study the problem of few-shot learning with the prism of inference on a partially observed graphical model, constructed from a collection of input images whose label can be either observed or not. By assimilating generic message-passing inference algorithms with their neural-network counterparts, we define a graph neural network architecture that generalizes several of the recently proposed few-shot learning models. Besides providing improved numerical performance, our framework is easily extended to variants of few-shot learning, such as semi-supervised or active learning, demonstrating the ability of graph-based models to operate well on ‘relational’ tasks.

AB - We propose to study the problem of few-shot learning with the prism of inference on a partially observed graphical model, constructed from a collection of input images whose label can be either observed or not. By assimilating generic message-passing inference algorithms with their neural-network counterparts, we define a graph neural network architecture that generalizes several of the recently proposed few-shot learning models. Besides providing improved numerical performance, our framework is easily extended to variants of few-shot learning, such as semi-supervised or active learning, demonstrating the ability of graph-based models to operate well on ‘relational’ tasks.

UR - http://www.scopus.com/inward/record.url?scp=85071163394&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85071163394&partnerID=8YFLogxK

M3 - Paper

ER -