MetAL: Active Semi-Supervised Learning on Graphs via Meta Learning

TL;DR

MetAL introduces a meta-learning based active semi-supervised learning method for graphs, selecting instances that most improve future classification performance, outperforming existing algorithms across various datasets.

Contribution

It formulates active learning on graphs as a bilevel optimization problem and employs meta-gradients to effectively select informative unlabeled instances.

Findings

MetAL outperforms state-of-the-art AL algorithms on multiple graph datasets.

Meta-gradient approximation effectively guides instance selection.

The approach improves classification accuracy with fewer labeled samples.

Abstract

The objective of active learning (AL) is to train classification models with less number of labeled instances by selecting only the most informative instances for labeling. The AL algorithms designed for other data types such as images and text do not perform well on graph-structured data. Although a few heuristics-based AL algorithms have been proposed for graphs, a principled approach is lacking. In this paper, we propose MetAL, an AL approach that selects unlabeled instances that directly improve the future performance of a classification model. For a semi-supervised learning problem, we formulate the AL task as a bilevel optimization problem. Based on recent work in meta-learning, we use the meta-gradients to approximate the impact of retraining the model with any unlabeled instance on the model performance. Using multiple graph datasets belonging to different domains, we demonstrate that MetAL efficiently outperforms existing state-of-the-art AL algorithms.