BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Hard and soft equivariance priors via Steerable CNNs - Dr Gabriele
  Cesa - Qualcomm AI Research\, Amsterdam
DTSTART:20250528T140500Z
DTEND:20250528T145500Z
UID:TALK229612@talks.cam.ac.uk
CONTACT:Ben Karniely
DESCRIPTION:Equivariance can enhance the data efficiency of machine learni
 ng models by incorporating prior knowledge about a problem.\nThanks to the
 ir flexibility and generality\, steerable CNNs are a popular design choice
  for equivariant networks.\nBy leveraging concepts from harmonic analysis\
 , these networks model symmetries through specific constraints on their le
 arnable weights or filters.\nThis framework facilitates the practical impl
 ementation of a wide variety of equivariant architectures - e.g. to most E
 uclidean isometries\, including E(3)\, E(2) and their subgroups.\n \nHowev
 er\, unknown or imperfect symmetries can sometimes lead to overconstrained
  weights and suboptimal performance.\nThis challenge motivated the study o
 f strategies to enforce softer priors into the models.\nIn the second half
  of this talk\, we will discuss a novel probabilistic approach to learning
  the degrees of equivariance in steerable CNNs.\nThe method replaces the e
 quivariance constraint on the weights with an expectation over a learnable
  distribution\, which is analytically computed by leveraging its Fourier d
 ecomposition.\n\n\nLink to join virtually: https://cam-ac-uk.zoom.us/j/874
 21957265\n\nA recording of this talk is available at the following link: h
 ttps://www.cl.cam.ac.uk/seminars/wednesday/video/\n\nThis talk is being re
 corded. If you do not wish to be seen in the recording\, please avoid sitt
 ing in the front three rows of seats in the lecture theatre. Any questions
  asked will also be included in the recording. The recording will be made 
 available on the Department’s webpage
LOCATION:Lecture Theatre 1\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR