GeomCA: Geometric Evaluation of Data Representations
Abstract
Evaluating the quality of learned representations without relying on a downstream task remains one of the challenges in representation learning. In this work, we present Geometric Component Analysis (GeomCA) algorithm that evaluates representation spaces based on their geometric and topological properties. GeomCA can be applied to representations of any dimension, independently of the model that generated them. We demonstrate its applicability by analyzing representations obtained from a variety of scenarios, such as contrastive learning models, generative models and supervised learning models.
Cite
Text
Poklukar et al. "GeomCA: Geometric Evaluation of Data Representations." International Conference on Machine Learning, 2021.Markdown
[Poklukar et al. "GeomCA: Geometric Evaluation of Data Representations." International Conference on Machine Learning, 2021.](https://mlanthology.org/icml/2021/poklukar2021icml-geomca/)BibTeX
@inproceedings{poklukar2021icml-geomca,
title = {{GeomCA: Geometric Evaluation of Data Representations}},
author = {Poklukar, Petra and Varava, Anastasiia and Kragic, Danica},
booktitle = {International Conference on Machine Learning},
year = {2021},
pages = {8588-8598},
volume = {139},
url = {https://mlanthology.org/icml/2021/poklukar2021icml-geomca/}
}