STATSTRO 2026 — Sampling, Simulation, and Scientific Discovery

Thank You

STATSTRO 2026 has come to a close. Over two days at the University of Toronto, our keynote speakers, tutorial leaders, contributed and lightning-talk presenters, poster authors, and attendees — joining us both in person and online from across the Great Lakes region, Canada, and around the world — came together to explore sampling, simulation, and scientific discovery.

To everyone who spoke, taught, presented, asked a sharp question, or simply shared an idea over coffee: thank you. You made this a warm, rigorous, and genuinely interdisciplinary gathering. The hands-on tutorial materials remain freely available, and we hope to see you at a future edition.

STATSTRO 2026 participants gathered for a group photo at the University of Toronto — STATSTRO 2026 participants · University of Toronto · July 16, 2026 · view full size

— The STATSTRO Organizing Committee

About the Workshop

STATSTRO (formerly Stellar Stats) is an annual interdisciplinary workshop that brings together statisticians, astronomers, members of the broader scientific community, and particularly early-career researchers working on interdisciplinary research in astrostatistics and astroinformatics.

Building on the Stellar Stats workshops (2021–2023) and previous STATSTRO editions (2024–2025), the workshop fosters connections between the Department of Statistical Sciences (DoSS), the Data Sciences Institute (DSI), and the astronomy communities of DADDAA, Dunlap, and CITA at the University of Toronto. This year's edition expanded the workshop into an international 2-day gathering, with a focus on communities around the Great Lakes area in the US and across Canada — though it was open to participants from anywhere in the world, both in person and remotely.

This year's theme, Sampling, Simulation, and Scientific Discovery, explores how modern sampling methods, computational simulations, and statistical inference are driving new discoveries across the sciences. The extended format allowed for longer tutorials, more in-depth discussions, stronger recruitment of statistics and machine learning researchers, and deeper engagement from both local and international participants.

Each of the four thematic sessions featured a keynote overview talk, a hands-on coding tutorial (Jupyter/Colab), and a contributed science presentation. Each day also brought a round of lightning talks from early-career researchers — each paired with a poster — followed by a dedicated poster session over the extended midday break. There was no registration fee, and catered meals and networking activities provided ample opportunities for cross-disciplinary connections.

Thematic Sessions

Deep Learning Day 1
Deep learning for scientific applications, including neural networks as emulators and surrogates, interpolation, extrapolation, and generalizability.
Uncertainty Quantification Day 1
Strategies for quantifying uncertainty across frequentist and Bayesian frameworks, including conformal prediction methods.
Sampling Techniques Day 2
Modern MCMC methods for scientific applications, with a focus on scaling to high dimensions and large datasets.
Simulation-Based Inference Day 2
Inference with intractable likelihoods using both traditional methods and neural approaches such as normalizing flows and diffusion models.

Speakers

Deep Learning

Ricardo Baptista

University of Toronto

Keynote Speaker

"Successes and Challenges of Posterior Sampling with Score-Based Diffusion Models"

Kartheik Iyer

Columbia University

Tutorial Leader

"How Do You Know What Your Models Are Learning?"

Tutorial materials

Ali SaraerToosi

University of Toronto

Contributed Speaker

"NeuralDMD: Interpretable Neural Representation of Dynamics from Sparse and Noisy Measurements"

Uncertainty Quantification

Mikael Kuusela

Carnegie Mellon University

Keynote Speaker

"Statistical Foundations of Uncertainty Quantification for Physicists in the Era of Machine Learning"

Biprateep Dey

University of Toronto

Tutorial Leader

"A Practitioner's Guide to Uncertainty Quantification"

Tutorial materials

Michael Evans

University of Toronto

Contributed Speaker

"Confidence, Statistical Evidence and Relative Belief with Applications to a Problem in Particle Physics"

Sampling Techniques

Radu Craiu

University of Toronto

Keynote Speaker

"The Universe of Sampling: Notes from a Statistical Odyssey"

Yichen Ji

University of Toronto

Tutorial Leader

"Bayesian Workflow Using PyMC"

Tutorial materials

Peter Behroozi

University of Arizona

Contributed Speaker

"The Ray Tracing Sampler: Bayesian Sampling of Neural Networks for Everyone"

Simulation-Based Inference

Justine Zeghal

Université de Montréal

Keynote Speaker

"From Simulations to Posteriors: A Tour of Simulation-Based Inference"

Connor Stone

University of Toronto

Tutorial Leader

"Building Forward Models for Astronomical Applications in the caskade Ecosystem"

Tutorial materials

Lawrence Faria

Queen's University

Contributed Speaker

"Simulation-Based Inference for HI Kinematics in Ultra-Diffuse Galaxies"

Schedule

All times in Eastern Daylight Time (EDT, UTC-4)

9:00 – 9:30

Registration & Setup

Check-in, coffee, and poster setup

9:30 – 9:40

Opening Remarks

9:40 – 10:40

Successes and Challenges of Posterior Sampling with Score-Based Diffusion Models

Keynote Talk

Ricardo Baptista — University of Toronto

Deep Learning

Keynote overview talk — 60 min

10:40 – 10:55

Break

10:55 – 11:40

How Do You Know What Your Models Are Learning?

Tutorial

Kartheik Iyer — Columbia University

Deep Learning

Hands-on coding tutorial — 45 min (Jupyter/Colab), delivered remotely

Tutorial materials

11:40 – 12:10

NeuralDMD: Interpretable Neural Representation of Dynamics from Sparse and Noisy Measurements

Contributed Talk

Ali SaraerToosi — University of Toronto

Deep Learning

Contributed science talk — 30 min

12:10 – 12:25

Lightning Talks

Lightning

8 × 1-minute lightning talks (5 min changeover) — a one-minute preview of every poster on display today

12:25 – 13:30

Lunch & Networking

Group photo, then catered lunch. Posters are up — take an extended break and start browsing before the dedicated session.

13:30 – 14:30

Poster Session

Poster

Dedicated poster viewing — meet the presenters behind today's lightning talks

14:30 – 15:30

Statistical Foundations of Uncertainty Quantification for Physicists in the Era of Machine Learning

Keynote Talk

Mikael Kuusela — Carnegie Mellon University

Uncertainty Quantification

Keynote overview talk — 60 min

15:30 – 15:45

Break

15:45 – 16:30

A Practitioner's Guide to Uncertainty Quantification

Tutorial

Biprateep Dey — University of Toronto

Uncertainty Quantification

Hands-on coding tutorial — 45 min (Jupyter/Colab)

Tutorial materials

16:30 – 17:00

Confidence, Statistical Evidence and Relative Belief with Applications to a Problem in Particle Physics

Contributed Talk

Michael Evans — University of Toronto

Uncertainty Quantification

Contributed science talk — 30 min

9:00 – 9:30

Registration & Setup

Check-in, coffee, and poster setup

9:30 – 10:30

The Universe of Sampling: Notes from a Statistical Odyssey

Keynote Talk

Radu Craiu — University of Toronto

Sampling Techniques

Keynote overview talk — 60 min

10:30 – 10:45

Break

10:45 – 11:30

Bayesian Workflow Using PyMC

Tutorial

Yichen Ji — University of Toronto

Sampling Techniques

Hands-on coding tutorial — 45 min (Jupyter/Colab)

Tutorial materials

11:30 – 12:00

The Ray Tracing Sampler: Bayesian Sampling of Neural Networks for Everyone

Contributed Talk

Peter Behroozi — University of Arizona

Sampling Techniques

Contributed science talk — 30 min

12:00 – 12:15

Lightning Talks

Lightning

7 × 1-minute lightning talks (5 min changeover) — a one-minute preview of every poster on display today

12:15 – 13:20

Lunch & Networking

Catered lunch. Posters are up — take an extended break and start browsing before the dedicated session.

13:20 – 14:20

Poster Session

Poster

Dedicated poster viewing — meet the presenters behind today's lightning talks

14:20 – 15:20

From Simulations to Posteriors: A Tour of Simulation-Based Inference

Keynote Talk

Justine Zeghal — Université de Montréal

Simulation-Based Inference

Keynote overview talk — 60 min

15:20 – 15:35

Break

15:35 – 16:20

Building Forward Models for Astronomical Applications in the caskade Ecosystem

Tutorial

Connor Stone — University of Toronto

Simulation-Based Inference

Hands-on coding tutorial — 45 min (Jupyter/Colab)

Tutorial materials

16:20 – 16:50

Simulation-Based Inference for HI Kinematics in Ultra-Diffuse Galaxies

Contributed Talk

Lawrence Faria — Queen's University

Simulation-Based Inference

Contributed science talk — 30 min

16:50 – 17:00

Closing Remarks

Wrap-up & next steps

Tutorial Materials

Each thematic session included a hands-on coding tutorial. The materials below were shared by the tutorial leaders and remain freely available to revisit and reuse.

Deep Learning

How Do You Know What Your Models Are Learning?

Kartheik Iyer — Columbia University

View materials

Uncertainty Quantification

A Practitioner's Guide to Uncertainty Quantification

Biprateep Dey — University of Toronto

View materials

Sampling Techniques

Bayesian Workflow Using PyMC

Yichen Ji — University of Toronto

View materials

Simulation-Based Inference

Building Forward Models for Astronomical Applications in the caskade Ecosystem

Connor Stone — University of Toronto

View materials

Posters & Lightning Talks

Each poster was previewed by a 1-minute lightning talk just before lunch, followed by a dedicated poster viewing session after lunch. Posters were grouped by thematic session and split across the two days.

Day 1 — Thursday, July 16 8 posters

Deep Learning

David Bromley

University of Toronto

“Spacetime Tomography via Learned Geodesics and Differentiable Rendering”

Nolan Koblischke

University of Toronto

“Vision-Language Models for Astrophysics”

Isabelle Laing

University of Toronto

“DOROTHY: A Stellar Catalogue for 13 Million Milky Way Stars from a Machine-Learning Pipeline”

Uncertainty Quantification

Alexandros Pratsos

University of Toronto

“Characterizing Stellar Streams with Error-Aware Machine Learning”

Benjamin Naeve Velguth

Dartmouth College

“Detecting Evidence of Hierarchical Structure Formation Around Dwarf Galaxies: Current and Future Observations”

Tanveer Karim

University of Toronto

“Is Dark Energy a Cosmological Constant? A Log Predictive Density Perspective”

Alexandra Rochon

McMaster University

“Understanding the Impact of Cold Gas Giants on the Formation of Super-Earths and Sub-Neptunes with Astrometry”

Megan Oxland

McMaster University

“Tracing Satellite Galaxy Evolution Across Cosmic Time”

Day 2 — Friday, July 17 7 posters

Sampling Techniques

Andrea Crespi

University of Waterloo

“Efficient Gradient-Based Sampling for Cosmological Field-Level Inference”

Mohan Agrawal

McGill University

“How to Generate Exact 1/fᵅ-Type Noise over an Arbitrary Number of Frequency Decades Without Running Out of Memory”

Bennett Neil Skinner

McMaster University

“Inferring Planet Compositions Using Statistical Methods”

Vincent Hénault-Brunet

Saint Mary's University

“Orbit-Based Constraints on the Mass and Position of an IMBH in Omega Centauri from Fast-Moving Stars”

Nasser Mohammed

University of Toronto

“Bayesian Mixture Modelling to Characterize Stellar Streams”

Simulation-Based Inference

Jennifer Y. H. Chan

Oberlin College

“Directional Multiscale Tools for Inference on the Sphere: From Wavelets to Curvelets with S2LET”

Callista Sullivan

Queen's University

“Can Simulation-Based Inference Reshape the Search for Structured Protostellar Disks?”

Thank you to all of our poster presenters and lightning-talk speakers for sharing their work across the two days.

Venue & Travel

Location

Department of Statistical Sciences

700 University Avenue, 9th–10th floors

Toronto, ON

WiFi

Visitors connected via eduroam. UofT guest WiFi credentials were provided at registration for those without eduroam access.

Tutorials

The hands-on coding sessions ran in Jupyter/Colab. The tutorial materials are now freely available to revisit.

Accessibility

The venue is wheelchair accessible.

Getting There

700 University Avenue is located in downtown Toronto, easily accessible by TTC subway (Queen's Park station on Line 1) or streetcar (College St or Dundas St).

Attendance & Logistics

STATSTRO 2026 has now taken place and registration is closed. Thank you to everyone who joined us — in person and online. See the recap and group photo and the tutorial materials above.

All participants were asked to follow our Code of Conduct.

In-Person & Remote Attendance

STATSTRO was a hybrid workshop with in-person and remote (Zoom) attendance. Registered participants received the Zoom link and logistics by email — the same link worked for both days.

Meals & Refreshments

Lunches on both days were catered and designed for networking. Snacks and beverages were provided during all breaks.

Conference Dinner (Day 1)

A conference dinner was held on Thursday, July 16 at 6:30 pm at Rikki Tikki (71 Jarvis St). Seats were allocated by lottery through the dinner sign-up form; confirmed guests received details by email.

Contributed Talks

Each thematic session featured a 30-minute contributed science talk from an invited researcher. See the schedule for this year's contributed speakers.

Tutorials

Each thematic session included a 45-minute hands-on coding tutorial (Jupyter/Colab). The tutorial materials shared by the leaders are now freely available.

Lightning Talks

Each day featured a round of rapid-fire 1-minute lightning talks from early-career researchers — one for each poster — held just before lunch as a preview of the day's posters. See the Posters & Lightning Talks section for the full lineup.

Poster Session

Every lightning talk had an accompanying poster, with a dedicated viewing session right after lunch each day. See the Posters & Lightning Talks section for who presented.