Talk @shapescience

Pip install topological data analysis

I’ve long wanted to do something real with TDA. There is something about the beautiful mathematical foundations, the efficiency of the algorithms, the oh-now-there-is-better-than-tSNE, the mystical visualisations that no one else understands (“just give me a bar chart with the mean”)…

There are some low hanging fruits to pick from the scikit-tda package. This umbrella package integrates with a scikitlearn API various TDA packages. For a lot of practical applications, it’s enough to get started. There are persistence diagrams, tools to turn them into features…

The package is young, but seems to keep improving. There were efforts lately to improve the docs, add notebooks, incorporate new libraries…

Now the big question is… how can you use this? On the top of my wishlist of the things I’ve been wanting to do with Mapper: feature influence when doing black-box optimization.