A Perception-Driven Approach to Supervised Dimensionality Reduction for Visualization

Abstract

Dimensionality reduction (DR) is a common strategy for visual analysis of labeled high-dimensional data. Low-dimensional representations of the data help, for instance, to explore the class separability and the spatial distribution of the data. Widely-used unsupervised DR methods like PCA do not aim to maximize the class separation, while supervised DR methods like LDA often assume certain spatial distributions and do not take perceptual capabilities of humans into account. These issues make them ineffective for complicated class structures. Towards filling this gap, we present a perception-driven linear dimensionality reduction approach that maximizes the perceived class separation in projections. Our approach builds on recent developments in perception-based separation measures that have achieved good results in imitating human perception. We extend these measures to be density-aware and incorporate them into a customized simulated annealing algorithm, which can rapidly generate a near optimal DR projection. We demonstrate the effectiveness of our approach by comparing it to state-of-the-art DR methods on 93 datasets, using both quantitative measure and human judgments. We also provide case studies with class-imbalanced and unlabeled data.

ClustMe: A Visual Quality Measure for Ranking Monochrome Scatterplots based on Cluster Patterns

“…Different options exist but finding the most efficient way to compute ClustMe is an important avenue for future work. As VQM for class‐separation have been used to guide supervised projection techniques [WFC*18], ClustMe could possibly be used as a criterion for Projection Pursuit [FT74].…”

Quality Metrics for Information Visualization

“…Thus, matrix reordering incorporates an implicit notion or understanding of the expected patterns. Another classical example in the context of dimension reduction is presented by Wang et al in [WFC*18]. For labeled datasets, typically depicted by color‐coded Scatter Plots, they start with a (pseudo‐)random placement of items in 2D.…”

“…In more advanced analysis scenarios, dimension reduction techniques are often used to map high‐dimensional features into 2D projection views [WFC*18]. For instance, principal component analysis is a projection technique that uses traditional Scatter Plots to map high‐dimensional data into a lower‐dimensional space [WEG87].…”

Interactive Correction of Mislabeled Training Data

“…Distance consistency difference (DSCD) To evaluate the performance of class structure preservation after sampling and projection, we adopt the density-aware distance consistency (DSC) [50], which measures the position of an item in its class. Given an item i, when its DSC value c i is close to 1, it is located close to the center of its class; when c i is close to -1, i is likely to located in wrong classes.…”