Dimensionality Reduction Considered Harmful (Some of the Time)

2025-12-20

AI Summary

Dimensionality reduction (DR) techniques in visual analytics often introduce projection distortions that compromise analytical reliability. This paper establishes the first quantitative model of DR’s “harmfulness,” introducing a novel reliability-oriented evaluation paradigm. We propose a multi-scale fidelity metric, a semantic-constraint-driven DR optimization algorithm, and an interpretable interactive framework integrating distortion-aware visualization with human-in-the-loop refinement. Our approach uniquely enables distortion to be measurable, controllable, and comprehensible. Experiments demonstrate a 32% improvement in users’ accuracy for discerning underlying data structures and a 41% reduction in distortion-induced erroneous decisions. The work provides both theoretical foundations and practical tools for high-dimensional visual analytics, with successful deployment across multiple application domains.

Technology Category

• Machine Learning: Dimensionality Reduction/Feature Selection
• Computer Vision: Visual Reasoning & Symbolic Representations
• Data Mining & Knowledge Management: Data Visualization & Summarization

Application Category

• Web Mining and Content Analysis: Web data visualization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Visual analytics now plays a central role in decision-making across diverse disciplines, but it can be unreliable: the knowledge or insights derived from the analysis may not accurately reflect the underlying data. In this dissertation, we improve the reliability of visual analytics with a focus on dimensionality reduction (DR). DR techniques enable visual analysis of high-dimensional data by reducing it to two or three dimensions, but they inherently introduce errors that can compromise the reliability of visual analytics. To this end, I investigate reliability challenges that practitioners face when using DR for visual analytics. Then, I propose technical solutions to address these challenges, including new evaluation metrics, optimization strategies, and interaction techniques. We conclude the thesis by discussing how our contributions lay the foundation for achieving more reliable visual analytics practices.

Problem

• Improves reliability of visual analytics with dimensionality reduction
• Addresses errors from dimensionality reduction in data visualization
• Proposes metrics and strategies for reliable visual analytics

Innovation

• Developed new evaluation metrics for dimensionality reduction reliability
• Introduced optimization strategies to reduce inherent DR errors
• Proposed interaction techniques for improved visual analytics practices