Seonglae ChoTDA
A methodology for analyzing data from a Topology perspective rather than coordinate values.
- Core Ideas:
- Data point cloud → construct complexes through neighborhood relationships
- Track when topological structures appear and disappear as the scale (radius) changes
(e.g., Vietoris–Rips complex, Čech complex)
- Representative Outputs:
- persistent homology
- persistence diagram / barcode
Why use it:
- Dimension-insensitive
- Relatively robust to noise
- Can numerically summarize the "shape essence" of data