Events / Statistics as linear algebra: distribution embeddings and operators on RKHS – Ingmar Schuster

Statistics as linear algebra: distribution embeddings and operators on RKHS – Ingmar Schuster

21st January 2021
14:00 - 15:00

Online

We are back for the new year with a thrilling new talk on kernel methods!
Join us on Thursday the 21st of January at 2pm UK time to find out more about what Ingmar Schuster’s research on Kernel Methods at Zalando Research looks like. You can read a synopsis for the talk below.

 

Statistics as linear algebra: distribution embeddings and operators on reproducing kernel Hilbert spaces
Embedding distribution into reproducing kernel Hilbert spaces (RKHSs) and expressing conditional distributions as operators is an idea that has been discussed extensively in the machine learning and statistics literature over the past 15 years. In this talk I will give an overview over my own and collaborators work in this area. This includes estimating densities and support from distribution embeddings, spectral analysis of operators for exploration of time series data, and finite sample statistical error bounds.

 

You can join the talk by clicking on the following ZOOM link or by using the Meeting ID 969 5789 4334. It will last 45 minutes + 15 minutes for questions and an informal chat.