Distribution Free Prediction Sets for Node Classification

Jase Clarkson

Keywords: Node Classification, Uncertainty Quantification, Conformal Prediction

TL;DR: We extend conformal prediction to graph structured data to construct valid prediction sets for inductive node classification tasks.

Abstract: Graph Neural Networks (GNNs) are able to achieve high classification accuracy on many large real world datasets, but provide no rigorous notion of predictive uncertainty. We leverage recent advances in conformal prediction to construct prediction sets for node classification in inductive learning scenarios, and verify the efficacy of our approach across standard benchmark datasets using popular GNN models. The code is available at \href{https://github.com/jase-clarkson/graph_cp}{this link}.

Type Of Submission: Extended abstract (max 4 main pages).

Agreement: Check this if you are okay with being contacted to participate in an anonymous survey.

Type Of Submission: Extended abstract.

Software: https://github.com/jase-clarkson/graph_cp

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2211.14555/code)