Augment to Interpret: Unsupervised and Inherently Interpretable Graph Embeddings

TL;DR

This paper introduces INGENIOUS, a framework for creating inherently interpretable graph embeddings through data augmentation, enabling better interpretability without post-hoc analysis while maintaining state-of-the-art performance.

Contribution

The paper presents INGENIOUS, a novel method for unsupervised, inherently interpretable graph embeddings that preserve semantics and improve downstream task performance.

Findings

Interpretable embeddings achieve state-of-the-art results

Data augmentation preserves semantics for interpretability

New metrics address interpretability in unsupervised learning

Abstract

Unsupervised learning allows us to leverage unlabelled data, which has become abundantly available, and to create embeddings that are usable on a variety of downstream tasks. However, the typical lack of interpretability of unsupervised representation learning has become a limiting factor with regard to recent transparent-AI regulations. In this paper, we study graph representation learning and we show that data augmentation that preserves semantics can be learned and used to produce interpretations. Our framework, which we named INGENIOUS, creates inherently interpretable embeddings and eliminates the need for costly additional post-hoc analysis. We also introduce additional metrics addressing the lack of formalism and metrics in the understudied area of unsupervised-representation learning interpretability. Our results are supported by an experimental study applied to both graph-level and node-level tasks and show that interpretable embeddings provide state-of-the-art performance on subsequent downstream tasks.