BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Maximilian Nickel (Facebook AI)
DTSTART:20220711T073000Z
DTEND:20220711T083000Z
DTSTAMP:20260424T221400Z
UID:GaML/1
DESCRIPTION:Title: Representation Learning and Generative Modeling on Manifolds\nby M
aximilian Nickel (Facebook AI) as part of Workshop on Geometry and Machine
Learning\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/GaML/1/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Nicolas Guigui (CNRS)
DTSTART:20220711T090000Z
DTEND:20220711T103000Z
DTSTAMP:20260424T221400Z
UID:GaML/2
DESCRIPTION:Title: Introduction to Geometric Statistics with Geomstats I\nby Nicolas
Guigui (CNRS) as part of Workshop on Geometry and Machine Learning\n\nAbst
ract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/GaML/2/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Beatrice Pozzetti (Heidelberg)
DTSTART:20220711T120000Z
DTEND:20220711T124000Z
DTSTAMP:20260424T221400Z
UID:GaML/3
DESCRIPTION:Title: Graph Embeddings in Symmetric Spaces\nby Beatrice Pozzetti (Heidel
berg) as part of Workshop on Geometry and Machine Learning\n\nAbstract: TB
A\n
LOCATION:https://stable.researchseminars.org/talk/GaML/3/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Pim de Haan (Amsterdam)
DTSTART:20220711T130000Z
DTEND:20220711T134000Z
DTSTAMP:20260424T221400Z
UID:GaML/4
DESCRIPTION:Title: Gauge Equivariant Mesh Convolutional Neural Networks\nby Pim de Ha
an (Amsterdam) as part of Workshop on Geometry and Machine Learning\n\n\nA
bstract\nConvolutional neural networks are widely successful in deep learn
ing on image datasets. However\, some data\, like that resulting from MRI
scans\, do not reside on a square grid\, but instead live on curved manifo
lds\, discretized as meshes. A key issue on such meshes is that they lack
a local notion of direction and hence the convolutional kernel cannot be c
anonically oriented. By doing message passing on the mesh and defining a g
roupoid of similar messages that should share weights\, we propose a gauge
equivariant method of building a CNN on such meshes that is direction-awa
re\, yet agnostic to how the directions are chosen. It is scalable\, invar
iant to how the mesh is rotated\, and performs state-of-the-art on a medic
al application for estimating blood flow through human arteries.\n
LOCATION:https://stable.researchseminars.org/talk/GaML/4/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Xavier Pennec (INRIA)
DTSTART:20220712T073000Z
DTEND:20220712T083000Z
DTSTAMP:20260424T221400Z
UID:GaML/5
DESCRIPTION:Title: Geometric Statistics for Computational Anatomy\nby Xavier Pennec (
INRIA) as part of Workshop on Geometry and Machine Learning\n\n\nAbstract\
nAt the interface of geometry\, statistics\, image analysis and medicine\,
computational anatomy aims at analysing and modelling the biological vari
ability of the organs shapes and their dynamics at the population level. T
he goal is to model the mean anatomy\, its normal variation\, its motion /
evolution and to discover morphological differences between normal and pa
thological groups. However\, shapes are usually described by equivalence c
lasses of sets of points\, curves\, surfaces or images under the action of
a transformation group\, or directly by the diffeomorphic deformation of
a template in diffeomorphometry. This implies that they live in non-linear
spaces\, while statistics where essentially developed in a Euclidean fram
ework. For instance\, adding or subtracting curves or surfaces does not re
ally make sense. Thus\, there is a need for redefining a consistent statis
tical framework for objects living in manifolds and Lie groups\, a field w
hich is now called geometric statistics. The objective of this talk is to
give an overview of the Riemannian computational tools and of simple stati
stics in these spaces. The talk is motivated and illustrated by applicatio
ns in medical image analysis\, such as the regression of simple and effici
ent models of the atrophy of the brain in Alzheimer’s disease and the gr
oupwise analysis of the motion of the heart in sequences of images using t
he parallel transport of surface and image deformations.\n
LOCATION:https://stable.researchseminars.org/talk/GaML/5/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Maxim Kochurov (PyMC Labs)
DTSTART:20220712T090000Z
DTEND:20220712T103000Z
DTSTAMP:20260424T221400Z
UID:GaML/6
DESCRIPTION:Title: Hyperbolic Manifolds in Deep Learning I\nby Maxim Kochurov (PyMC L
abs) as part of Workshop on Geometry and Machine Learning\n\n\nAbstract\nH
yperbolic manifolds are quite new in deep learning. Mathematical elegance
and theoretical advantages are very attractive properties for dimensionali
ty reduction and rich representations. Moreover\, a lot of research was do
ne to investigate opportunities in graph-based deep learning or language m
odels. In the talk I’ll give an overview of what are the main advances i
n the area\, highlighting the most problematic theory and motivation. Duri
ng the practical session\, we’ll get familiar with models and implementa
tions that make use of the hyperbolic space to their fullest potential.\n
LOCATION:https://stable.researchseminars.org/talk/GaML/6/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Björn Ommer (LMU Munich)
DTSTART:20220712T120000Z
DTEND:20220712T124000Z
DTSTAMP:20260424T221400Z
UID:GaML/7
DESCRIPTION:by Björn Ommer (LMU Munich) as part of Workshop on Geometry a
nd Machine Learning\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/GaML/7/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Erik Bekkers (Amsterdam)
DTSTART:20220712T130000Z
DTEND:20220712T134000Z
DTSTAMP:20260424T221400Z
UID:GaML/8
DESCRIPTION:Title: Geometric and Physical Quantities improve E(3) Equivariant Message Pas
sing\nby Erik Bekkers (Amsterdam) as part of Workshop on Geometry and
Machine Learning\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/GaML/8/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Emanuele Rodolà (Sapienza University)
DTSTART:20220713T073000Z
DTEND:20220713T083000Z
DTSTAMP:20260424T221400Z
UID:GaML/9
DESCRIPTION:Title: From Sound to Metric Priors: A New Paradigm for Shape Generation\n
by Emanuele Rodolà (Sapienza University) as part of Workshop on Geometry
and Machine Learning\n\n\nAbstract\nSpectral and metric geometry are at th
e heart of various problems in computer vision\, graphics\, pattern recogn
ition\, and machine learning. Ultimately\, the core reason for their succe
ss can be traced down to questions of stability and to the informativeness
of the eigenvalues of certain operators. In this talk\, I will discuss an
d show tangible examples of such properties and showcase some dramatic imp
lications on a selection of notoriously hard problems in computer vision a
nd graphics. First\, I will address the question of whether one can recove
r the shape of a geometric object from its vibration frequencies (‘hear
the shape of the drum’)\; while theoretically the answer to this questio
n is negative\, little is known about the practical possibility of using t
he spectrum for shape reconstruction and optimization. I will introduce a
numerical procedure called isospectralization\, as well as a data-driven v
ariant\, showing how this *practical* problem is solvable. Then\, I will d
iscuss the increasingly popular task of designing an effective generative
model for deformable 3D shapes. I will demonstrate how injecting metric di
stortion priors into a simple geometric reconstruction loss can lead to th
e formation of a very informative latent space\, which can be trained with
extremely scarce data (less than 10 examples) and still yield competitive
generation quality as well as aiding geometric disentanglement.\n
LOCATION:https://stable.researchseminars.org/talk/GaML/9/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Anastasis Kratsios (McMaster)
DTSTART:20220713T090000Z
DTEND:20220713T094000Z
DTSTAMP:20260424T221400Z
UID:GaML/10
DESCRIPTION:Title: Embedding Guarantees for Representations by Small Probabilistic Graph
Transformers\nby Anastasis Kratsios (McMaster) as part of Workshop on
Geometry and Machine Learning\n\nAbstract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/GaML/10/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Nicolas Guigui (CNRS)
DTSTART:20220713T110000Z
DTEND:20220713T123000Z
DTSTAMP:20260424T221400Z
UID:GaML/11
DESCRIPTION:Title: Introduction to Geometric Statistics with Geomstats II\nby Nicola
s Guigui (CNRS) as part of Workshop on Geometry and Machine Learning\n\nAb
stract: TBA\n
LOCATION:https://stable.researchseminars.org/talk/GaML/11/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Maxim Kochurov (PyMC Labs)
DTSTART:20220713T130000Z
DTEND:20220713T143000Z
DTSTAMP:20260424T221400Z
UID:GaML/12
DESCRIPTION:Title: Hyperbolic Manifolds in Deep Learning II\nby Maxim Kochurov (PyMC
Labs) as part of Workshop on Geometry and Machine Learning\n\n\nAbstract\
nHyperbolic manifolds are quite new in deep learning. Mathematical eleganc
e and theoretical advantages are very attractive properties for dimensiona
lity reduction and rich representations. Moreover\, a lot of research was
done to investigate opportunities in graph-based deep learning or language
models. In the talk I’ll give an overview of what are the main advances
in the area\, highlighting the most problematic theory and motivation. Du
ring the practical session\, we’ll get familiar with models and implemen
tations that make use of the hyperbolic space to their fullest potential.\
n
LOCATION:https://stable.researchseminars.org/talk/GaML/12/
END:VEVENT
END:VCALENDAR