Dear all,

There is an online seminar today by Sanjoy Dasgupta https://cseweb.ucsd.edu/~dasgupta/ at 15.00, which may be of broader interest. Sanjoy is internationally well known for his work on clustering and unsupervised learning, and has a reputation as an excellent speaker. His talk today will be about new results on interpretable clustering. See below for details.

This is part of a new online seminar about Theory for Interpretable AI. For further seminar info, see tverven.github.io/tiai-seminar/ http://tverven.github.io/tiai-seminar/

Best,   Tim

-------- Forwarded Message -------- Subject: [TIAI-seminar] Sanjoy Dasgupta, July 11 Date: Fri, 5 Jul 2024 21:37:36 +0300 From: Michal Moshkovitz via TIAI-seminar tiai-seminar@list.uva.nl Reply-To: Michal Moshkovitz michal.moshkovitz@mail.huji.ac.il To: tiai-seminar@list.uva.nl

Dear all,

Unsupervised learning is a significant branch of machine learning, with k-means clustering being a key subproblem. However, it often lacks interpretability. Can we make it interpretable? Apparently we can.

On this topic, our next speaker in the Theory of Interpretable AI Seminar, Sanjoy Dasgupta, will review recent progress on interpretable k-means clustering.

*Speaker*: Sanjoy Dasgupta https://cseweb.ucsd.edu/~dasgupta/ *Date*: Thursday, July 11, 15.00 Central European Time (CET) / 9.00 am Eastern Standard Time (EST) *Zoom link*:https://uva-live.zoom.us/j/87120549999

*Title*: Recent progress on interpretable clustering

The widely-used k-means procedure returns k clusters that have arbitrary convex shapes. In high dimension, such a clustering might not be easy to understand. A more interpretable alternative is to constrain the clusters to be the leaves of a decision tree with axis-parallel splits; then each cluster is a hyper-rectangle given by a small number of features. Is it always possible to find clusterings that are interpretable in this sense and yet have k-means cost that is close to the unconstrained optimum? A recent line of work has answered this in the affirmative and moreover shown that these interpretable clusterings are easy to construct. I will give a survey of these results: algorithms, methods of analysis, and open problems.

*General Seminar info*: Web:tverven.github.io/tiai-seminar/ http://tverven.github.io/tiai-seminar/ Google calendar:Google calendarhttps://calendar.google.com/calendar/u/1?cid=NTlhNjNhZDQ5ZmUxYmM5MmRmZTMwNzkwOWZhYjMyNTRhMzA4OGYwZTAxY2Q5MGU3NzQ2YjRlNWE0NzhmMzFkMUBncm91cC5jYWxlbmRhci5nb29nbGUuY29t

*Upcoming speakers*: - September 5: Lesia Semenova - October 10: Ulrike von Luxburg

Best Regards, Michal Moshkovitz

Also on behalf of TIAI co-organizers Suraj Srinivas and Tim van Erven