AI Seminar: Gauge Equivariant CNNs and coordinate independent convolutions
Maurice Weiler, PhD Researcher from University of Amsterdam, Netherlands.
Gauge Equivariant CNNs and coordinate independent convolutions
The idea of equivariance to symmetry transformations provides one of the first theoretically grounded principles for neural network architecture design.
In this talk we focus on equivariant convolutional networks for processing spatiotemporally structured signals like images, audio or configurations of physical systems.
We start by demanding equivariance w.r.t. global symmetries of the underlying space and proceed by generalizing the resulting design principles to local gauge transformation, thereby enabling the development of equivariant convolutional networks on general manifolds.
Defining its feature spaces as spaces of coordinate independent feature fields, the theory of Gauge Equivariant Convolutional Networks shows intriguing parallels with fundamental theories in physics.
Beyond unifying several lines of research in equivariant and geometric deep learning, Gauge Equivariant Convolutional Networks are demonstrated to yield greatly improved performances compared to conventional CNNs in a wide range of signal processing tasks.
The seminar is free and open for everyone.
This seminar is a part of the AI Seminar Series organised by SCIENCE AI Centre. The series highlights advances and challenges in research within Machine Learning, Data Science, and AI. Like the AI Centre itself, the seminar series has a broad scope, covering both new methodological contributions, ground-breaking applications, and impacts on society.