Clustering-Based Relational Unsupervised Representation Learning with an Explicit Distributed Representation

TL;DR

This paper introduces a novel unsupervised relational representation learning method that models data as hypergraphs, using clustering to generate features, leading to improved classification performance and lower complexity.

Contribution

It presents a new hypergraph-based clustering approach for relational data, enhancing unsupervised representation learning with broad similarity measures.

Findings

Models on the new representations outperform existing methods.

The approach achieves lower complexity in relational tasks.

Experimental results demonstrate improved classification accuracy.

Abstract

The goal of unsupervised representation learning is to extract a new representation of data, such that solving many different tasks becomes easier. Existing methods typically focus on vectorized data and offer little support for relational data, which additionally describe relationships among instances. In this work we introduce an approach for relational unsupervised representation learning. Viewing a relational dataset as a hypergraph, new features are obtained by clustering vertices and hyperedges. To find a representation suited for many relational learning tasks, a wide range of similarities between relational objects is considered, e.g. feature and structural similarities. We experimentally evaluate the proposed approach and show that models learned on such latent representations perform better, have lower complexity, and outperform the existing approaches on classification tasks.