Attendees sit and engage at a conference presentation during CVPR (Conference on Computer Vision and Pattern Recognition) in Seattle. The large conference room is filled with rows of chairs, occupied by participants attentively watching the ongoing session. The presentation topic, visible on the large screen at the front, is titled "Gromov-Wasserstein for Encoding Structural Priors," indicating a focus on mathematical formulations and cost matrices in the context of computer vision problems. The venue is equipped with an industrial design, featuring an open ceiling with visible beams and ducts, as well as high-tech equipment. Among the attendees, one individual captures the moment on a smartphone, highlighting the blend of modern technology and academic engagement. The atmosphere is one of focused learning and knowledge exchange.
Text transcribed from the image:
omov-Wasserstein for
coding Structural Priors
atively) general formulation for (discrete) GW problems:
minimize ELC.CT
CVPR
JUNE 17-21, 2024
MIT SEATTLE, W
Gromov-Wasserstein for
Encoding Structural Priors
A (relatively) general t
ete) GW problems
st
Tix N
TINIK
t matrices C" e RN and C
a" function between cost i
Cost matric
"Loss" fur
nts &: RXR-R
Gromov-Wasserstein for
Encoding Structural Priors
A (relatively) general formulation for (discrete) GW problems:
minimize Σike[N] L(CC)TiTkl
TERNXK
s.t.
j,le[K]
T1K = N,
T1N=1K,
> Cost matrices C" E RNXN and Ca Є RKXK
Australian
National
University
>"Loss" function between cost matrix elements L: RX R-R
MegXu and Shephen Gould Semporaly Content Undanced Optimal Transport for Unsupervised