BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Edward Pearce-Crump (Imperial College London\, UK) DTSTART:20240520T140000Z DTEND:20240520T150000Z DTSTAMP:20260422T181228Z UID:QGS/115 DESCRIPTION:Title: Co mpact Matrix Quantum Group Equivariant Neural Networks\nby Edward Pear ce-Crump (Imperial College London\, UK) as part of Quantum Groups Seminar [QGS]\n\n\nAbstract\nIn deep learning\, we would like to develop principle d approaches for constructing neural networks. One important approach invo lves identifying symmetries that are inherent in data and then encoding th em into neural network architectures using representations of groups. Howe ver\, there exist so-called “quantum symmetries” that cannot be unders tood formally by groups. In this talk\, we show how to construct neural ne tworks that are equivariant to compact matrix quantum groups using Woronow icz’s version of Tannaka-Krein duality. We go on to characterise the lin ear weight matrices that appear in these neural networks for a class of co mpact matrix quantum groups known as “easy”. In particular\, we show that every compact matrix group equivariant neural network is a compact ma trix quantum group equivariant neural network.\n LOCATION:https://researchseminars.org/talk/QGS/115/ END:VEVENT END:VCALENDAR