Abstract
Machine learning methods were historically developed for Euclidean spaces with fixed dimension, a prime example being tabular data and digital images of fixed size. In my industrial career, I have noticed that enterprises store data in structured formats, like JSONs, XML, ProtoBuffers, etc. It can be safe to say that structured data formats are powering the internet as we know it today. Seeing the gap, in this talk, I will talk about methods we have developed to fill this gap. Thanks to the right level of abstraction, strong theoretical guarantees, and good performance, the methods have been successfully applied in the antivirus industry, protecting 400M+ users.
13.05.2026, 13.00 - 14.00, T2:D3-209, prezenční formou
