The Challenges of Automating Clustering

The paper is out. So if you want to read it and then try to implement it, I want to write some code for it. It's not too hard. You just summarize each partition. The choice of summary statistics matters, but we list that in the paper. You cluster the summary statistics, so you identify correlated features. And then you just do a nearest neighbors on your clusters. If the cluster is found anomalous, then those are your broken features.

Transcript

Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!

Get the app