A Quality Metric to Improve Scatterplots for Explainable AI
Loading...
Date
2024
Journal Title
Journal ISSN
Volume Title
Publisher
The Eurographics Association
Abstract
Scatterplots are widely utilised in Explainable Artificial Intelligence (XAI) to investigate misclassifications and patterns among instances. However, when datasets are large, overplotting diminishes the effectiveness of scatterplots. This poster introduces a new quality metric to measure the overplotting of scatterplots in the context of XAI. Initially, we assess the significance of each data point within a scatterplot by continuous density transformation, Mahalanobis Distance and a mapping function. Building on this foundation, we develop a quality metric for scatterplots. Our metric performs well accounting for rendering orders and marker sizes in scatterplots, showcasing the metric's potential to improve the effectiveness of XAI scatterplots.
Description
CCS Concepts: Human-centered computing → Visualisation design and evaluation methods; Computing methodologies → Machine learning
@inproceedings{10.2312:evp.20241077,
booktitle = {EuroVis 2024 - Posters},
editor = {Kucher, Kostiantyn and Diehl, Alexandra and Gillmann, Christina},
title = {{A Quality Metric to Improve Scatterplots for Explainable AI}},
author = {Liu, Liqun and Ruddle, Roy A. and Bogachev, Leonid V. and Rezaei, Mahdi and Khara, Arjun},
year = {2024},
publisher = {The Eurographics Association},
ISBN = {978-3-03868-258-5},
DOI = {10.2312/evp.20241077}
}