Metadata Representations for Queryable Repositories of Machine Learning Models

Journal Article (2023)
Author(s)

Z. Li (TU Delft - Web Information Systems)

Henk Kant (Student TU Delft)

Rihan Hai (TU Delft - Web Information Systems)

Asterios Katsifodimos (TU Delft - Web Information Systems)

Marco Brambilla (Politecnico di Milano)

Alessandro Bozzon (TU Delft - Sustainable Design Engineering)

Research Group
Web Information Systems
Copyright
© 2023 Z. Li, Henk Kant, R. Hai, A Katsifodimos, Marco Brambilla, A. Bozzon
DOI related publication
https://doi.org/10.1109/ACCESS.2023.3330647
More Info
expand_more
Publication Year
2023
Language
English
Copyright
© 2023 Z. Li, Henk Kant, R. Hai, A Katsifodimos, Marco Brambilla, A. Bozzon
Research Group
Web Information Systems
Volume number
11
Pages (from-to)
125616-125630
Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

Machine learning (ML) practitioners and organizations are building model repositories of pre-trained models, referred to as model zoos. These model zoos contain metadata describing the properties of the ML models and datasets. The metadata serves crucial roles for reporting, auditing, ensuring reproducibility, and enhancing interpretability. Despite the growing adoption of descriptive formats like datasheets and model cards, the metadata available in existing model zoos remains notably limited. Moreover, existing formats have limited expressiveness, thus constraining the potential use of model repositories, extending their purpose beyond mere storage for pre-trained models. This paper proposes a unified metadata representation format for model zoos. We illustrate that comprehensive metadata enables a diverse range of applications, encompassing model search, reuse, comparison, and composition of ML models. We also detail the design and highlight the implementation of an advanced model zoo system built on top of our proposed metadata representation.