Topological Metric for Unsupervised Embedding Quality Evaluation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Shestov, Aleksei, Klenitskiy, Anton, Denisova, Daria, Dzagkoev, Amurkhan, Petrovich, Daniil, Savchenko, Andrey, Makarenko, Maksim
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908718059225088
author Shestov, Aleksei
Klenitskiy, Anton
Denisova, Daria
Dzagkoev, Amurkhan
Petrovich, Daniil
Savchenko, Andrey
Makarenko, Maksim
author_facet Shestov, Aleksei
Klenitskiy, Anton
Denisova, Daria
Dzagkoev, Amurkhan
Petrovich, Daniil
Savchenko, Andrey
Makarenko, Maksim
contents Modern representation learning increasingly relies on unsupervised and self-supervised methods trained on large-scale unlabeled data. While these approaches achieve impressive generalization across tasks and domains, evaluating embedding quality without labels remains an open challenge. In this work, we propose Persistence, a topology-aware metric based on persistent homology that quantifies the geometric structure and topological richness of embedding spaces in a fully unsupervised manner. Unlike metrics that assume linear separability or rely on covariance structure, Persistence captures global and multi-scale organization. Empirical results across diverse domains show that Persistence consistently achieves top-tier correlations with downstream performance, outperforming existing unsupervised metrics and enabling reliable model and hyperparameter selection.
format Preprint
id arxiv_https___arxiv_org_abs_2512_15285
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Topological Metric for Unsupervised Embedding Quality Evaluation
Shestov, Aleksei
Klenitskiy, Anton
Denisova, Daria
Dzagkoev, Amurkhan
Petrovich, Daniil
Savchenko, Andrey
Makarenko, Maksim
Machine Learning
Information Retrieval
Modern representation learning increasingly relies on unsupervised and self-supervised methods trained on large-scale unlabeled data. While these approaches achieve impressive generalization across tasks and domains, evaluating embedding quality without labels remains an open challenge. In this work, we propose Persistence, a topology-aware metric based on persistent homology that quantifies the geometric structure and topological richness of embedding spaces in a fully unsupervised manner. Unlike metrics that assume linear separability or rely on covariance structure, Persistence captures global and multi-scale organization. Empirical results across diverse domains show that Persistence consistently achieves top-tier correlations with downstream performance, outperforming existing unsupervised metrics and enabling reliable model and hyperparameter selection.
title Topological Metric for Unsupervised Embedding Quality Evaluation
topic Machine Learning
Information Retrieval
url https://arxiv.org/abs/2512.15285