Topological Data Analysis
summary
Topological Data Analysis converts a dataset and a chosen distance into an overlapping-bin projection, clusters each bin, then links clusters that share points to produce a network model that highlights distinct hotspots. Applied to a bank stress-test dataset it reduces thousands of indicators to four representative features; in healthcare it finds outlier care paths that cut length-of-stay. Shape is measured with persistent homology (persistence barcodes used as features), and explanations use Kolmogorov–Smirnov-ranked variables.
AI-written summary · about
chapters (10)
- 0:00Introduction and context
- 1:51Meet Gunnar Carlsson
- 3:35Introduction to TDA
- 7:07Case study on model explainability
- 10:43Technical mechanics of TDA
- 17:22Interacting with TDA models
- 21:05Healthcare applications
- 25:32Resources for further learning
- 28:30Persistent homology and deep learning
- 33:05Closing remarks
related talks
50mThe Shape of Data: Machine Learning and Topology
06mComplex Data Analysis
43mTopological data analysis: How to quantify "holes" in your data and why?
51mTopological space creation and Clustering at BigData scale
28mAnthony Bak, Principal Data Scientist at Ayasdi @ MLconf SEA
20mAllison Gilmore, Data Scientist, Ayasdi
40mTopological Techniques for Unsupervised Learning
26mMathematical Approaches to Clustering