A Topological Heat Map for Data Analysis (TopHeat)
Abstract
Topological data analysis provides summaries for the shape of data. Some of these summaries, such as the PI's persistence landscape, are feature maps and kernels, and can be easily combined with standard methods of statistics and machine learning. However, these topological summaries can be difficult for non-experts to interpret. In this project, we produced a new summary, that may be visualized as a heat map on the underlying data. We developed theory for this heat map, showing that it is stable under perturbations of the input. Furthermore, we showed how to combine this summary with statistics and machine learning and applied it to synthetic data and real data.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 05, 2023
- Accession Number
- AD1224710
Entities
Organizations
- University of Florida