analiza aiqbench
This research formalizes the "Rashomon effect" in dimensionality reduction, demonstrating that multiple valid embeddings can coexist while preserving data structure. The authors propose a framework leveraging this multiplicity to enhance interpretability and robustness through PCA alignment and concept regularization. By extracting persistent nearest-neighbor relationships across these solutions, the method yields more trustworthy visualizations for high-dimensional datasets.
The Rashomon Effect for Visualizing High-Dimensional Data
arXiv:2604.00485v3 Announce Type: replace 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' embedding -- 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.
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' embedding -- 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.
Machine Learning (cs.LG)
arXiv:2604.00485 [cs.LG]
(or arXiv:2604.00485v3 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2604.00485 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 247 cuvinte
Subiecte conexe
Hybrid quantum-neural network beats classical machine learning
Quantum Zeitgeist · 17 aug. 2026
On the Diffusibility of High-Dimensional Latents
arXiv · 24 sept. 2026
STRIPS-WM: Learning Grounded Propositional STRIPS-style World Models from Images
arXiv · 8 iun. 2026
MAPS: A Novel Multi-Axial Projective Sphere for Geometrically Visualizing Higher d-Valued Quantum State-Space of Qudits
arXiv · 17 iun. 2026