MECpace: A Visual Analytics Tool for Comparing Multiple Embedding Spaces
Loading...
Date
2025
Journal Title
Journal ISSN
Volume Title
Publisher
The Eurographics Association
Abstract
Embeddings play a crucial role in machine learning (ML) by representing high-dimensional data in a lower-dimensional space, enhancing model efficiency and interoperability. Fine-grained analysis of embeddings enables optimization of model architectures, refinement of datasets, and more effective parameter adjustments. We introduce MECpace (Multiple Embedded Comparison Spaces), a web-based visualization tool designed to facilitate the comparison of multiple embedding spaces through intuitive visualizations. MECpace supports rapid comparisons of object neighbors across models using parallel coordinate plots and scatter plots, as well as pairwise comparisons through an integrated matrix-scatter view that combines scatter plots and histograms. Interactive features such as filtering and zooming enable seamless exploration of large datasets. By providing a comprehensive view of embedding similarities across multiple models, MECpace enhances decision-making in the ML pipeline.
Description
CCS Concepts: Human-centered computing → Visualization; Computing methodologies → Machine learning
@inproceedings{10.2312:evp.20251128,
booktitle = {EuroVis 2025 - Posters},
editor = {Diehl, Alexandra and Kucher, Kostiantyn and Médoc, Nicolas},
title = {{MECpace: A Visual Analytics Tool for Comparing Multiple Embedding Spaces}},
author = {Joshi, Rachit and Zinjarde, Purva and Humayoun, Shah Rukh},
year = {2025},
publisher = {The Eurographics Association},
ISBN = {978-3-03868-286-8},
DOI = {10.2312/evp.20251128}
}