Update paper (#519)

Co-authored-by: manzt <[email protected]>

_publications/manz-cev-2024.md

@@ -1,40 +1,22 @@
 ---
-title: "A General Framework for Comparing Embedding Visualizations Across Class-Label Hierarchies"
+title: A General Framework for Comparing Embedding Visualizations Across Class-Label Hierarchies
 image: cev.png
 members:
   - trevor-manz
   - fritz-lekschas
   - nils-gehlenborg
 year: 2024
 type: article
-publisher: "https://ieeexplore.ieee.org/document/10672535/"
-doi: "10.1109/TVCG.2024.3456370"
-zotero-key: "EBDFTS7Q"
-preprint: "https://doi.org/10.31219/osf.io/puxnf"
+publisher: 'https://ieeexplore.ieee.org/document/10672535/'
+doi: 10.1109/TVCG.2024.3456370
+zotero-key: EBDFTS7Q
+preprint: 'https://doi.org/10.31219/osf.io/puxnf'
 cite:
-  authors: "T Manz, F Lekschas, E Greene, G Finak, N Gehlenborg"
-  published: "*IEEE Transactions on Visualization and Computer Graphics* 1-11"
+  authors: 'T Manz, F Lekschas, E Greene, G Finak, N Gehlenborg'
+  published: '*IEEE Transactions on Visualization and Computer Graphics* 1-11'
+videos: []
+other-resources: []
+awards: []
+code: 'https://github.com/OzetteTech/comparative-embedding-visualization'
 ---
-Projecting high-dimensional vectors into two dimensions for visualization,
-known as embedding visualization, facilitates perceptual reasoning and
-interpretation. Comparison of multiple embedding visualizations drives
-decision-making in many domains, but conventional comparison methods are
-limited by a reliance on direct point correspondences. This requirement
-precludes embedding comparisons without point correspondences, such as two
-different datasets of annotated images, and fails to capture meaningful
-higher-level relationships among point groups. To address these shortcomings,
-we propose a general framework to compare embedding visualizations based on
-shared class labels rather than individual points. Our approach partitions
-points into regions corresponding to three key class concepts--confusion,
-neighborhood, and relative size--to characterize intra- and inter-class
-relationships. Informed by a preliminary user study, we realize an
-implementation of our framework using perceptual neighborhood graphs to define
-these regions and introduce metrics to quantify each concept. We demonstrate
-the generality of our framework with use cases from machine learning and
-single-cell biology, highlighting our metrics' ability to draw insightful
-comparisons across label hierarchies. To assess the effectiveness of our
-approach, we conducted a user study with five machine learning researchers and
-six single-cell biologists using an interactive and scalable prototype
-developed in Python and Rust. Our metrics enable more structured comparison
-through visual guidance and increased participants’ confidence in their
-findings.
+Projecting high-dimensional vectors into two dimensions for visualization, known as embedding visualization, facilitates perceptual reasoning and interpretation. Comparison of multiple embedding visualizations drives decision-making in many domains, but conventional comparison methods are limited by a reliance on direct point correspondences. This requirement precludes embedding comparisons without point correspondences, such as two different datasets of annotated images, and fails to capture meaningful higher-level relationships among point groups. To address these shortcomings, we propose a general framework to compare embedding visualizations based on shared class labels rather than individual points. Our approach partitions points into regions corresponding to three key class concepts--confusion, neighborhood, and relative size--to characterize intra- and inter-class relationships. Informed by a preliminary user study, we realize an implementation of our framework using perceptual neighborhood graphs to define these regions and introduce metrics to quantify each concept. We demonstrate the generality of our framework with use cases from machine learning and single-cell biology, highlighting our metrics' ability to draw insightful comparisons across label hierarchies. To assess the effectiveness of our approach, we conducted a user study with five machine learning researchers and six single-cell biologists using an interactive and scalable prototype developed in Python and Rust. Our metrics enable more structured comparison through visual guidance and increased participants’ confidence in their findings.