Few-Shot Learning with Graph Neural Networks

TL;DR

This paper introduces a graph neural network framework for few-shot learning that leverages message-passing inference on graphical models, improving performance and extending to semi-supervised and active learning scenarios.

Contribution

It presents a novel GNN architecture that unifies and generalizes existing few-shot learning models through inference on graphical models.

Findings

Improved numerical performance over existing models

Flexible extension to semi-supervised and active learning

Effective operation on relational tasks

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.