一种用于估计可视化收益的有限测度
人工智能
2022-04-21 v2 图形学
人机交互
信息论
math.IT
摘要
信息论可用于分析可视化过程的成本-收益。然而,当前的收益测度包含一个无界项,既不易估计也不直观可解释。在本工作中,我们提出通过用有界项替换无界项来修正现有的成本-收益测度。我们考察了若干有界测度,包括 Jensen-Shannon 散度以及作为本工作一部分而提出的一种新散度测度。我们使用视觉分析来支持多准则比较,将搜索范围缩小到具有更好数学性质的那几个选项。我们将剩余选项应用于两个可视化案例研究,以在具体场景中实例化其用法,而收集到的真实世界数据进一步为有限测度的选择提供了依据,该测度可用于估计可视化的收益。
引用
@article{arxiv.2002.05282,
title = {A Bounded Measure for Estimating the Benefit of Visualization},
author = {Min Chen and Mateu Sbert and Alfie Abdul-Rahman and Deborah Silver},
journal= {arXiv preprint arXiv:2002.05282},
year = {2022}
}
备注
Comment on version 2: This revised version, which includes a new formal proof, many additions, and a detailed revision report, was submitted to SciVis 2020. Unexpectedly, our revision effort did not have much influence on the SciVis 2020 reviewers who gave an outright rejection with lower scores than EuroVis reviews. We will share these reviews after we have completed our feedback