Deepvoc: a linked open vocabulary for reproducible and reliable deep learning experiments

Document Type

Article

Source of Publication

International Journal of Machine Learning and Cybernetics

Publication Date

6-13-2025

Abstract

Several Deep Learning (DL) algorithms and techniques have been developed and published in recent years to address problems across various domains. To ensure accurate result comparisons, DL experiments must be conducted in a consistent computing environment using the same algorithm configurations and datasets. Developing DL algorithms requires programmers to manage numerous parameter settings and datasets, making it challenging to test and document results accurately without proper metadata provenance. However, the connected data community lacks a lightweight metadata exchange framework for DL across different environments, limiting high-level interoperability. This article bridges that gap by introducing DeepVoc (Deep Learning Vocabulary) a structured vocabulary designed to enhance data provenance and improve the usability of DL experiments. Moreover, DeepVoc adheres to the Findable, Accessible, Interoperable, and Reusable (FAIR) principles, ensuring better metadata standardization while enhancing the reproducibility, reusability, and interoperability of DL experiments.

ISSN

1868-8071

Publisher

Springer Science and Business Media LLC

Disciplines

Computer Sciences

Keywords

Deep learning, Experiments, FAIR, Interoperability, Metadata, Reusability

Scopus ID

105007854819

Indexed in Scopus

yes

Open Access

no

Share

COinS