The Rashomon Effect for Visualizing High-Dimensional Data

Yiyang Sun, Haiyang Huang, Gaurav Rajesh Parikh, Cynthia Rudin
Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, PMLR 300:3160-3168, 2026.

Abstract

Dimension reduction (DR) is inherently non-unique: multiple embeddings can preserve the structure of high-dimensional data equally well while differing in layout or geometry. In this paper, we formally define the Rashomon set for DR—the collection of ‘good’ embeddings—and show how embracing this multiplicity leads to more powerful and trustworthy representations. Specifically, we pursue three goals. First, we introduce PCA-informed alignment to steer embeddings toward principal components, making axes interpretable without distorting local neighborhoods. Second, we design concept-alignment regularization that aligns an embedding dimension with external knowledge, such as class labels or user-defined concepts. Third, we propose a method to extract common knowledge across the Rashomon set by identifying trustworthy and persistent nearest-neighbor relationships, which we use to construct refined embeddings with improved local structure while preserving global relationships. By moving beyond a single embedding and leveraging the Rashomon set, we provide a flexible framework for building interpretable, robust, and goal-aligned visualizations.

Cite this Paper

BibTeX
@InProceedings{pmlr-v300-sun26c, title = { The Rashomon Effect for Visualizing High-Dimensional Data }, author = {Sun, Yiyang and Huang, Haiyang and Parikh, Gaurav Rajesh and Rudin, Cynthia}, booktitle = {Proceedings of The 29th International Conference on Artificial Intelligence and Statistics}, pages = {3160--3168}, year = {2026}, editor = {Khan, Emtiyaz and Li, Yingzhen and Solin, Arno and Ramdas, Aaditya}, volume = {300}, series = {Proceedings of Machine Learning Research}, month = {02--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v300/main/assets/sun26c/sun26c.pdf}, url = {https://proceedings.mlr.press/v300/sun26c.html}, abstract = { Dimension reduction (DR) is inherently non-unique: multiple embeddings can preserve the structure of high-dimensional data equally well while differing in layout or geometry. In this paper, we formally define the Rashomon set for DR—the collection of ‘good’ embeddings—and show how embracing this multiplicity leads to more powerful and trustworthy representations. Specifically, we pursue three goals. First, we introduce PCA-informed alignment to steer embeddings toward principal components, making axes interpretable without distorting local neighborhoods. Second, we design concept-alignment regularization that aligns an embedding dimension with external knowledge, such as class labels or user-defined concepts. Third, we propose a method to extract common knowledge across the Rashomon set by identifying trustworthy and persistent nearest-neighbor relationships, which we use to construct refined embeddings with improved local structure while preserving global relationships. By moving beyond a single embedding and leveraging the Rashomon set, we provide a flexible framework for building interpretable, robust, and goal-aligned visualizations. } }
Endnote
%0 Conference Paper %T The Rashomon Effect for Visualizing High-Dimensional Data %A Yiyang Sun %A Haiyang Huang %A Gaurav Rajesh Parikh %A Cynthia Rudin %B Proceedings of The 29th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2026 %E Emtiyaz Khan %E Yingzhen Li %E Arno Solin %E Aaditya Ramdas %F pmlr-v300-sun26c %I PMLR %P 3160--3168 %U https://proceedings.mlr.press/v300/sun26c.html %V 300 %X Dimension reduction (DR) is inherently non-unique: multiple embeddings can preserve the structure of high-dimensional data equally well while differing in layout or geometry. In this paper, we formally define the Rashomon set for DR—the collection of ‘good’ embeddings—and show how embracing this multiplicity leads to more powerful and trustworthy representations. Specifically, we pursue three goals. First, we introduce PCA-informed alignment to steer embeddings toward principal components, making axes interpretable without distorting local neighborhoods. Second, we design concept-alignment regularization that aligns an embedding dimension with external knowledge, such as class labels or user-defined concepts. Third, we propose a method to extract common knowledge across the Rashomon set by identifying trustworthy and persistent nearest-neighbor relationships, which we use to construct refined embeddings with improved local structure while preserving global relationships. By moving beyond a single embedding and leveraging the Rashomon set, we provide a flexible framework for building interpretable, robust, and goal-aligned visualizations.
APA
Sun, Y., Huang, H., Parikh, G.R. & Rudin, C.. (2026). The Rashomon Effect for Visualizing High-Dimensional Data . Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 300:3160-3168 Available from https://proceedings.mlr.press/v300/sun26c.html.

Related Material