Laurence Perreault-Levasseur

alexandre.adam@mila.quebec

Biographie

Laurence Perreault-Levasseur est titulaire de la Chaire de recherche du Canada en cosmologie computationnelle et en intelligence artificielle. Elle est professeure adjointe à l'Université de Montréal et membre associée de Mila – Institut québécois d’intelligence artificielle, où elle mène des recherches sur le développement et l'application de méthodes d'apprentissage automatique à la cosmologie. Elle est également chercheuse invitée au Flatiron Institute, à New York. Auparavant, elle a été chargée de recherche au Center for Computational Astrophysics du Flatiron Institute et boursière postdoctorale du KIPAC à l'Université de Stanford. Laurence Perreault-Levasseur a obtenu un doctorat de l'Université de Cambridge, où elle a travaillé sur les applications des méthodes de la théorie des champs effectifs ouverts au formalisme de l'inflation. Elle est titulaire d'une licence et d'une maîtrise en sciences de l'Université McGill.

Étudiants actuels

Alexandre Adam

Doctorat - UdeM

tara.akhoundsadegh@mila.quebec

Tara Akhound-Sadegh

Doctorat - McGill

Superviseur⋅e principal⋅e :

Siamak Ravanbakhsh

Missael Barco

Doctorat - UdeM

Co-superviseur⋅e :

missael-gabriel.barco@mila.quebec

Misha Barth

Doctorat - UdeM

Superviseur⋅e principal⋅e :

michael.barth@mila.quebec

mohammed.chamma@mila.quebec

Mohammed Chamma

Postdoctorat - McGill

Noé Dia

Stagiaire de recherche - UdeM

Co-superviseur⋅e :

Doctorat - UdeM

Co-superviseur⋅e :

andreas.filipp@umontreal.ca

Ronan Legin

Doctorat - UdeM

Superviseur⋅e principal⋅e :

ronan.legin@mila.quebec

christian.lemay1@sympatico.ca

Christian Lemay

Stagiaire de recherche - UdeM

Pablo Lemos

Postdoctorat - UdeM

Co-superviseur⋅e :

michael.matesic@mila.quebec

pablo.lemos@mila.quebec

Doctorat - UdeM

guillaume.payeur@mail.mcgill.ca

Guillaume Payeur

Doctorat - UdeM

Maîtrise recherche - UdeM

Superviseur⋅e principal⋅e :

nicolas.payot@mila.quebec

Sacha Perry-Fagant

Doctorat - UdeM

Co-superviseur⋅e :

sacha.perry-fagant@mila.quebec

sammy.sharief@mila.quebec

Salma Salhi

Doctorat - UdeM

salhi.salma@mila.quebec

Ketan Sand

Postdoctorat - McGill

ketan.sand@mila.quebec

Sammy Nasser Sharief

Maîtrise recherche - UdeM

Nolan Smyth

Postdoctorat - UdeM

Superviseur⋅e principal⋅e :

nolan.smyth@mila.quebec

Yitian Sun

Postdoctorat - McGill

Co-superviseur⋅e :

yitian.sun@mila.quebec

Justine Zeghal

Postdoctorat - UdeM

Co-superviseur⋅e :

justine.zeghal@mila.quebec

Laurence Perreault-Levasseur

Publications

GaMPEN: A Machine Learning Framework for Estimating Bayesian Posteriors of Galaxy Morphological Parameters

Aritra Ghosh

C. M. Urry

Amrit Rau

Miles Cranmer

Kevin Schawinski

Dominic Stark

Chuan Tian

Ryan Ofman

Tonima Tasnim Ananna

Connor Auge

N. Cappelluti

D. B. Sanders

Ezequiel Treister

We introduce a novel machine-learning framework for estimating the Bayesian posteriors of morphological parameters for arbitrarily large num… (voir plus)bers of galaxies. The Galaxy Morphology Posterior Estimation Network (GaMPEN) estimates values and uncertainties for a galaxy’s bulge-to-total-light ratio (L B /L T ), effective radius (R e ), and flux (F). To estimate posteriors, GaMPEN uses the Monte Carlo Dropout technique and incorporates the full covariance matrix between the output parameters in its loss function. GaMPEN also uses a spatial transformer network (STN) to automatically crop input galaxy frames to an optimal size before determining their morphology. This will allow it to be applied to new data without prior knowledge of galaxy size. Training and testing GaMPEN on galaxies simulated to match z 0.25 galaxies in Hyper Suprime-Cam Wide g-band images, we demonstrate that GaMPEN achieves typical errors of 0.1 in L B /L T , 0.″17 (∼7%) in R e , and 6.3 × 104 nJy (∼1%) in F. GaMPEN's predicted uncertainties are well calibrated and accurate (5% deviation)—for regions of the parameter space with high residuals, GaMPEN correctly predicts correspondingly large uncertainties. We a

2022-07-31

The Astrophysical Journal (publié)

doi.org

arxiv.org

Bayesian Neural Networks

Tom Charnock

Laurence Perreault-Levasseur

François Lanusse

In recent times, neural networks have become a powerful tool for the analysis of complex and abstract data models. However, their introducti… (voir plus)on intrinsically increases our uncertainty about which features of the analysis are model-related and which are due to the neural network. This means that predictions by neural networks have biases which cannot be trivially distinguished from being due to the true nature of the creation and observation of data or not. In order to attempt to address such issues we discuss Bayesian neural networks: neural networks where the uncertainty due to the network can be characterised. In particular, we present the Bayesian statistical framework which allows us to categorise uncertainty in terms of the ingrained randomness of observing certain data and the uncertainty from our lack of knowledge about how data can be created and observed. In presenting such techniques we show how errors in prediction by neural networks can be obtained in principle, and provide the two favoured methods for characterising these errors. We will also describe how both of these methods have substantial pitfalls when put into practice, highlighting the need for other statistical techniques to truly be able to do inference when using neural networks.

2022-02-05

Artificial Intelligence for High Energy Physics (publié)

doi.org

arxiv.org

Machine Learning Advantages in Canadian Astrophysics

Kim Venn

Sébastien Fabbro

Adrian Liu