BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:SDSC-AI4Science seminar: Advancing cosmology with deep learning
DTSTART:20230322T160000
DTEND:20230322T170000
DTSTAMP:20260511T213611Z
UID:63d77a1afbe1f2de682086e5343140add9a2b5645c02ced504012ff4
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr. Tomasz Kacprzak and Dr. Nathanaël Perraudin\, Sr. Data Sc
 ientists at SDSC\nPlace: GA3 21 (Bernoulli Center)\, followed by an apér
 o in GA3 31. (Zoom: https://epfl.zoom.us/j/67864899458)\n\nSpeaker: Dr. 
 Tomasz Kacprzak and Dr. Nathanaël Perraudin\, Sr. Data Scientists at SDSC
 \n\nTitle: Advancing cosmology with deep learning\n\nAbstract: Cosmology\
 , despite its seemingly disparate nature from machine learning\, is actual
 ly a data-driven field that can greatly benefit from recent advancements i
 n machine learning techniques. In this presentation\, we will showcase a c
 ollaborative SDSC project that exemplifies how machine learning can contri
 bute to significant breakthroughs in cosmology.\n\nIn the first part of th
 e talk\, we will give an overview the types of innovation that we applied 
 to cosmology: cosmological parameter measurements from dark matter maps us
 ing deep learning\, accelerating cosmological simulations using Generative
  Adversarial Networks\, and analysing data on the sphere using graph convo
 lutional neural networks.\n\nIn the second part of the talk\, we will do a
  deep-dive into the problem of machine learning on the sphere using graphs
  and present our approach called DeepSphere. Graphs are versatile data str
 uctures that can represent pairwise relationships between objects or act a
 s a discrete representation of a continuous manifold. Using the graph-base
 d representation\, we define many of the standard CNN operations\, such as
  convolution and pooling. With filters restricted to having radial symmetr
 y\, our convolutions are equivariant to rotation on the sphere\, and DeepS
 phere can be made invariant or equivariant to rotation. This way\, DeepSph
 ere is a special case of a graph CNN\, tailored to the HEALPix sampling of
  the sphere. This approach is computationally more efficient than using sp
 herical harmonics to perform convolutions. We demonstrate the method on a 
 classification problem of weak lensing mass maps from two cosmological mod
 els and compare its performance with that of three baseline classifiers\, 
 two based on the power spectrum and pixel density histogram\, and a classi
 cal 2D CNN. Our experimental results show that the performance of DeepSphe
 re is always superior or equal to the baselines.\n\nThis research project 
 was performed in close collaboration with the Cosmology Research Group of 
 ETH Zürich.
LOCATION:GA 3 21 https://plan.epfl.ch/?room==GA%203%2021 https://epfl.zoom
 .us/j/67864899458
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR