Portrait de Yashar Hezaveh

Yashar Hezaveh

Membre académique associé
yashar.hezaveh@mila.quebec
Professeur adjoint, Université de Montréal, Département de physique
Sujets de recherche
Apprentissage de représentations
Apprentissage profond
Vision par ordinateur
Site web
Google Scholar
LinkedIn

Biographie

Yashar Hezaveh est membre associé de Mila – Institut québécois d'intelligence artificielle et directeur de Ciela – Institut de Montréal pour l'analyse des données astrophysiques et l'apprentissage automatique. Il est professeur adjoint au Département de physique de l'Université de Montréal, titulaire d'une chaire de recherche du Canada en analyse de données astrophysiques et apprentissage automatique, membre associé de l'Institut spatial Trottier de l'Université McGill et chercheur invité au Center for Computational Astrophysics du Flatiron Institute (New York) et au Perimeter Institute. Auparavant, il a été chercheur au Flatiron Institute (2018-2019) et boursier Hubble de la NASA à l'Université de Stanford (2013-2018).

Il est un leader mondial dans l'analyse des données astrophysiques avec l'apprentissage automatique. Ses recherches actuelles portent principalement sur l'inférence bayésienne dans l'IA (par exemple, les modèles de diffusion) et visent à faire progresser les connaissances sur la distribution de la matière noire dans les galaxies fortement lenticulaires à l'aide de données provenant de grands relevés cosmologiques. Ses recherches sont soutenues par la Schmidt Futures Foundation et la Simons Foundation.

Étudiants actuels

Publications

Do Tabular Foundation Models Know Physics? Contamination, Units, and the Deterministic Limit
Wassim Tenachi
Yashar Hezaveh
Laurence Perreault Levasseur
Pierre-Luc Bacon
Tabular foundation models (TFMs) learn to fill in tables the way language models fill in text, and tables are arguably the format in which m… (voir plus)ost physical measurement arrives. Did they learn any physics in the process? They are Bayesian by construction, so the question is what their prior contains. We probe it directly, evaluating four of them (TabPFN-3, TabICLv2, TabDPT and Real-TabPFN-2.5) against six baselines on datasets sampled from 316 physical equations, in and out of domain. TFMs dominate, out of the box and after tuning. But we show that their prior can represent neither a noiseless mechanism nor physical units, which is why they interpolate physics without yet being able to act as physical models.
2026-12-30
OpenReview (publié)
doi.org
openreview.net
Microcaustics: Fast GPU-Accelerated Dynamical Microlensing of Time-Dependent Extended Sources
Joshua Fagin
Connor Stone
James Hung-Hsu Chan
Sophia Miskiewicz
Henry Best
Paras Sharma
Matthew O’Dowd
Laurence Perreault-Levasseur
Yashar Hezaveh
Simon Birrer
Gravitational microlensing probes unresolved quasar accretion disc structure and compact objects in lens galaxies and is an important system… (voir plus)atic in strong-lens time-delay cosmography. Most analyses use static microlensing map simulations because evolving individual microlenses and caustics is computationally prohibitive. We present Microcaustics, a GPU-accelerated Python package built on PyTorch for dynamical microlensing of time-dependent extended sources in strong gravitational lenses. It combines a local-exact/far-field deflection solver, source-directed lens-plane scouting with temporal reuse, and biquadratic inverse polygon mapping (IPM). GPU batching and fused Triton rasterization accelerate the calculation, while binary caustic labels track source-centred crossings. Across three simulated quadruply lensed quasar systems, our package generates a 10-year, 147-frame multiband light curve in an amortized median of 0.68 s (0.88 s with labels) on a consumer GPU with a median light-curve RMSE of 0.35 mmag, while recovering every reference source-centred caustic crossing. At comparable map accuracy, the batched throughput is 780 times higher than that of a recent CUDA IPM implementation and extrapolates to a speedup of millions relative to classic GPU-D. We demonstrate that our IPM converges as
2026-10-13
arXiv (prépublication)
doi.org
arxiv.org
PM++: Scaling Differentiable Cosmological Simulators with Dynamic Particle Ownership
Nicolas Payot
Yashar Hezaveh
Laurence Perreault-Levasseur
Large differentiable particle-mesh simulations support cosmological inference and simulator-coupled learning, but their evolving states can … (voir plus)exceed single-accelerator memory. Distribution with fixed particle ownership remains memory-limited because deposition halos must cover the maximum cumulative displacement. We present
2026-10-02
Sim2Sci @ Neural Information Processing Systems (publié)
openreview.net
openreview.net
POSTELLAR: Posterior Stellar Spectrum Sampling—An Alternative to Approximate Stellar Spectra for Exoplanetary Analysis
Dhvani Doshi
Nicolas B. Cowan
Yashar Hezaveh
Gabriel Missael Barco
Étienne Artigau
Abstract We present a novel approach to perform posterior sampling of the underlying stellar spectrum in high-resolution spectroscopic obser… (voir plus)vations. Our method, postellar , coherently combines information from empirical observations and physics-based models, enabling more accurate spectral recovery while providing uncertainties that can be propagated into downstream analyses. This is accomplished by treating the intrinsic stellar spectrum as a latent variable and performing posterior sampling under a Gaussian likelihood with an informative prior constructed using a score-based diffusion model trained on PHOENIX stellar models. We validate the framework on synthetic SPectropolarimètre InfraRouge (SPIRou) radial velocity (RV) observations generated from the empirical spectra of Barnard’s Star and Proxima Centauri. Spectra inferred with postellar recover the ground truth more accurately than standard empirical templates. For medium signal-to-noise observations, postellar improves RV accuracy by up to a factor of three, while RVs derived from empirical templates tend to be biased. This method performs particularly well in low signal-to-noise and low-cadence regimes. The postellar framework is broadly applicable to other high-resolution spectroscopy science cases, including stellar abundance analyses and exoplanet atmospheric characterization.
2026-09-13
Astrophysical Journal (publié)
doi.org
arxiv.org
Observation-driven simulations of strong lensing galaxy clusters
L. Leuzzi
Massimo Meneghetti
A. Adam
Lauro Moscardini
C. Giocoli
P. Bergamini
Yashar Hezaveh
Matteo Maturi
Amata Mercurio
A. Moretti
Andrés Alejandro Plazas Malagón
P. Rosati
Galaxy clusters are the most powerful strong lenses: they greatly magnify the flux of distant and faint sources. Strong lensing also allows … (voir plus)for the reconstruction of their mass distribution with-1% level accuracy and enables investigating cosmological parameters. The number of known systems of this type is bound to increase in the next years thanks to wide imaging surveys. Using simulations in this context is crucial to validate the analysis methods before they are applied to real data, and to train machine learning algorithms that can handle large volumes of images. In this work, we present a simulated set of one hundred images of galaxy clusters that we have produced with a novel code for simulating cluster-scale strong lenses. One of the main novelties of our approach, distinguishing it from other existing codes, is the use of empirical relations, derived from state-of-the-art observations, for modelling the characteristics, such as morphology, color, and spatial distribution of the cluster member population. This allows us to reliably reproduce the complexity of real observations. The simulations are partly carried out with the latest version of SkyLens, a code that creates mock observations of strong lensing events in different systems and observational setups. The main improvements we introduce are the use of the message passage interface (MPI) paradigm, a standard for parallel programming that leverages the use of several processors to perform some given task, and the implementation of score-based diffusion models to augment the images of the background sources. Together, they lead to more efficient and realistic image simulations. We also present the validation of the code and simulations by comparing the properties of the mock clusters to those of real ones. [Abridged]
2026-09-06
arXiv (prépublication)
doi.org
arxiv.org
Microlensing Detection and Inference via Learned Bayes Factors
Nolan Smyth
Laurence Perreault-Levasseur
Yashar Hezaveh
We present a unified framework for gravitational microlensing event detection and parameter inference. Traditional pipelines use determinist… (voir plus)ic hard cuts on photometric statistics, systematically missing low-magnification events in the finite-source regime. We instead frame detection as Bayesian model comparison using Evidence Networks, which learn calibrated Bayes factors from binary-labeled simulations, and combine this with Neural Posterior Estimation (NPE) for amortized parameter inference. Both share a transformer encoder that handles irregularly-sampled time series without imputation. On simulated Roman Space Telescope data, our Evidence Network achieves
2026-05-29
AI4Physics @ International Conference on Machine Learning (publié)
openreview.net
openreview.net
Strong Gravitational Lensing Posterior Sampling in Pixel-Space Using Diffusion Models and Recurrent Inference Machines
Guillaume Payeur
Laurence Perreault-Levasseur
Gabriel Missael Barco
Yashar Hezaveh
Modeling galaxy-galaxy strong gravitational lenses to infer the brightness of the source galaxy and the mass distribution of the foreground … (voir plus)galaxy is computationally challenging, particularly for high-resolution, high signal-to-noise observations. In this regime, high-dimensional representations of both the source and the foreground mass distribution are necessary to model the data down to the noise level. This inference problem has been challenging for both traditional and machine learning-based methods because of its high dimensionality and its non-linearity in the foreground mass distribution. We present a method to generate joint posterior samples of the source galaxy and foreground mass distribution as pixelated images conditioned on observations. The method combines diffusion-based generative modeling and recurrent inference machines. It can model realistic gravitational lensing simulations with background and foreground galaxies drawn from cosmological hydrodynamical simulations down to the noise level.
2026-05-29
AI4Physics @ International Conference on Machine Learning (publié)
openreview.net
openreview.net
MIRA: A Score for Conditional Distribution Accuracy and Model Comparison
Sammy Sharief
Justine Zeghal
Gabriel Missael Barco
Pablo Lemos
Yashar Hezaveh
Laurence Perreault-Levasseur
We present Mira, a method for estimating the expected probability that samples from a candidate conditional distribution match the true, unk… (voir plus)nown conditional distribution, for which only data-label pairs are available. We derive theoretical bounds obtained when the candidate distribution matches the true one and when the conditional distributions are independent. This framework thus enables model comparison by quantifying the alignment between the conditional distribution of a candidate model and the data-label pairs of the true model. Consequently, Mira enables Bayesian model comparison through direct posterior validation, bypassing the challenging evidence computation. We demonstrate its effectiveness across several toy problems and Bayesian inference tasks.
2026-04-29
International Conference on Machine Learning (spotlight)
doi.org
proceedings.mlr.press
LSST Strong Lensing Systems Dark Matter Sensitivity Analysis with Neural Ratio Estimators
Andreas Filipp
Yashar Hezaveh
Laurence Perreault-Levasseur
Daniel Gilman
LSST Dark Energy Science Collaboration
Strong gravitational lensing offers a unique probe of dark matter (DM) on sub-galactic scales, where the abundance and distribution of low-m… (voir plus)ass halos are highly sensitive to the underlying properties of DM particles. In this work, we forecast LSST's sensitivity to DM substructure in galaxy-galaxy strong lenses using simulated samples and neural ratio estimators (NREs). Our simulations include both subhalos within the main deflector and line-of-sight (LOS) halos, with halo masses down to
2026-04-07
arXiv (prépublication)
doi.org
arxiv.org
Caustics: A Python Package for Accelerated Strong Gravitational Lensing Simulations
Connor Stone
Alexandre Adam
Adam Coogan
M. J. Yantovski-Barth
Andreas Filipp
Landung Setiawan
Cordero Core
Ronan Legin
Charles Wilson
Gabriel Missael Barco
Yashar Hezaveh
Laurence Perreault-Levasseur
2025-12-31
Zenodo (publié)
doi.org
arxiv.org
Transformer Embeddings for Fast Microlensing Inference
Nolan Smyth
Laurence Perreault-Levasseur
Yashar Hezaveh
2025-12-11
ArXiv (prépublication)
arxiv.org
arxiv.org
Neural Deprojection of Galaxy Stellar Mass Profiles
M. J. Yantovski-Barth
Hengyue Zhang
Nolan Smyth
Connor Stone
Martin Bureau
Yashar Hezaveh
Laurence Perreault-Levasseur
We introduce a neural approach to dynamical modeling of galaxies that replaces traditional imaging-based deprojections with a differentiable… (voir plus) mapping. Specifically, we train a neural network to translate Nuker profile parameters into analytically deprojectable Multi Gaussian Expansion components, enabling physically realistic stellar mass models without requiring optical observations. We integrate this model into SuperMAGE, a differentiable dynamical modelling pipeline for Bayesian inference of supermassive black hole masses. Applied to ALMA data, our approach finds results consistent with state-of-the-art models while extending applicability to dust-obscured and active galaxies where optical data analysis is challenging.
2025-11-24
arXiv (prépublication)
doi.org
arxiv.org