Week 09 — Machine Learning with TDA
Where TDA features earn their place in a pipeline — and where they don't.
Week 09 — Machine Learning with TDA
Where TDA features earn their place in a pipeline — and where they don't.
Lecture
TDA features as ML inputs · concatenation vs replacement of standard features · when TDA wins (3D structure, sensor data, time series with cycles) and when it doesn’t (raw images, tabular data without geometric structure) · honest reporting of TDA gains.
Read before the lecture
- Hensel, Moor, Rieck, *A Survey of Topological Machine Learning Methods* (Front. Artif. Intell. 2021)
Problem set
PS5 — Replication exercise
- Replicate one TDA-ML claim from the literature on a public dataset. Report the gain (or lack thereof) over a non-TDA baseline you build.
Reference text for this week: chapter 09 of the bilingual notes — EN PDF · FR PDF.