SolarView: Low Distortion Radial Embedding with a Focus.

Area of Science:

• Computer Science
• Data Visualization
• Information Visualization

Background:

• High-dimensional data visualization presents challenges in preserving relationships and minimizing distortion.
• Existing dimensionality reduction techniques often struggle with maintaining accurate distances, especially for specific entities of interest.

Purpose of the Study:

• To introduce a novel radial embedding technique focused on preserving distances to a central entity.
• To develop an interactive tool, SolarView, for exploring high-dimensional data using this embedding.
• To compare the proposed embedding method against state-of-the-art techniques in terms of distortion and effectiveness.

Main Methods:

• A novel radial embedding algorithm was designed, prioritizing near-exact distances to a focus entity.
• The SolarView tool was created, featuring a "solar system" layout with concentric circles for neighbor embedding.
• The new embedding and adapted state-of-the-art methods were implemented and experimentally evaluated.
• Interaction techniques were incorporated into SolarView to aid users in judging visualization distortion.

Main Results:

• The proposed radial embedding method demonstrates competitive performance against state-of-the-art techniques.
• The embedding achieves low distortion in practice, particularly when adhering to the solar system design.
• Experimental evaluation confirmed the method's ability to preserve distances to the focus entity effectively.
• Interaction techniques provide clear visual cues for assessing distortion levels.

Conclusions:

• The novel radial embedding approach offers a viable alternative for visualizing high-dimensional data, especially when focusing on a central entity.
• SolarView provides an effective interactive environment for exploring complex datasets like bibliographic similarities.
• While effective, users should be aware of potential distortion inherent in dimensionality reduction, which SolarView helps to mitigate through interaction.