SUDS project

Oct 02 2026

Robustness and transparency for machine learning models

Research description:

The ability to quickly and accurately identify covariate shift at test time is a critical and often overlooked component of safe machine learning systems deployed in high-risk domains. While methods exist for detecting when predictions should not be made on out-of-distribution test examples, identifying distributional level differences between training and test time can help determine when a model should be removed from the deployment setting and retrained. This project will evaluate the Detectron model https://github.com/rgklab/detectron on a wide variety of datasets from the WILDS benchmark (https://wilds.stanford.edu/) and the SUBPOPBench (https://github.com/YyzHarry/SubpopBench) benchmark dataset. This will enable ML researchers to identify promising next steps to build guardrails to protect against distribution shift. The student will be responsible for coding, designing and running experiments using Pytorch on a large scale GPU cluster to study, compare and contrast different methods to detect when a machine learning model might fail on publicly available datasets. This project will introduce the student to slurm, pytorch and empirical research in machine learning.

Year: 2024

Researcher:
Rahul Krishnan, Department of Computer Science, Faculty of Arts & Science, University of Toronto

Students:
Siddharth Arya, University of Toronto
Mohamed Khalid Aljudaibi, King Abdullah University of Science & Technology

Through SUDS, undergraduate students engage in hands-on research focused on data sciences and AI methodology applications.

Learn more

Oct 02 2026

Searching for Stellar Streams with 10 Million Stars in the Milky Way

Research description:

Our galaxy, the Milky Way, is surrounded by numerous small galaxies and star clusters that can be influenced by its gravitational forces, leading to the formation of stellar streams— celestial “rivers” orbiting around our galaxy. These streams offer a unique opportunity for astronomers to delve into the mysteries of galaxy formation and the elusive nature of dark matter. (For an intriguing example, check out our feature in The Globe & Mail)

Thanks to cutting-edge cosmic surveys, we now have access to comprehensive data on millions of stars in our universe, including their full 6D information (position and velocity). The SUDS Scholar will be at the forefront of developing a Bayesian framework to assess the membership probability of each star in potential streams and to characterize the properties of these stellar streams. This involves leveraging vast astronomical datasets, totaling several gigabytes of data, obtained from one of the largest spectroscopic surveys, the Dark Energy Spectroscopic Instrument (DESI).

In this research project, the SUDS Scholar will explore the development and application of innovative statistical and computational techniques. These methodologies are crucial not only for unraveling the secrets hidden within stellar streams but also for paving the way for future astronomical surveys.

Year: 2024

Researcher:
Li Ting, David A. Dunlap Department of Astronomy and Astrophysics, Faculty of Arts & Science, University of Toronto

Student:
Joseph Tang, University Of Toronto Mississauga

Through SUDS, undergraduate students engage in hands-on research focused on data sciences and AI methodology applications.

Learn more

Oct 02 2026

The Impact of Solid Organ Transplantation on Work and Earnings

Research description:

Solid organ transplantation is a life-changing intervention for individuals with end-stage kidney, liver, pancreas, lung, or heart failure. Numerous studies have shown that transplantation improves survival and quality of life for recipients. However, as the number of transplants performed in Canada continues to increase, greater attention is needed towards understanding other outcomes important to patients, such as ability to work and earn income. There is currently little known about the labour market implications of solid organ transplantation. Previous studies in this area have been survey-based which can be biased due to small sample sizes and self-reported data. We propose to conduct a population-based retrospective cohort study examining income and employment of individuals who have undergone solid organ transplantation. We will leverage a unique dataset that our team has created called the Canadian Hospitalization and Taxation Database. Income, employment, and health data of all solid organ transplant patients in Canada will be derived from a linkage between the Canadian Institute for Health Information Discharge Abstract Database and the T1 Family File. Analysis will be conducted using advanced econometric and epidemiological methods. This study will be the first of its kind and have a significant impact on the field of transplant medicine.

The student will be expected to conduct a background literature search, develop a protocol, analyze data and prepare a manuscript for publication in a peer-reviewed journal for which they will be the first-author. The student will also attend weekly meetings with the research team and provide regular updates on the progress of the project.

Year: 2024

Researcher:
Karim Ladha, Unity Health Toronto

Students:
Daisy Thomas, University of Toronto
Shouq Alsulami, King Abdullah University of Science & Technology

Through SUDS, undergraduate students engage in hands-on research focused on data sciences and AI methodology applications.

Learn more

Oct 02 2026

The link between neuroinflammation and neurological markers of depression

Research description:

With a life-time prevalence of 12% in Canada, major depression is the third highest cause of disability worldwide. Mid-to-late-life depression (MLD) confers a 2-5 fold increase in dementia risk, including for Alzheimer’s disease. Also, it has become increasingly clear that depression has a strong link to neuroinflammation. Thus we emphasize the importance of grounding the study in the link between a neurological signature of depression and its link with inflammation. Furthermore, there are sex differences in the immune system. Men and women may also differ in the types of inflammatory markers they produce as they age. The premier approach to studying the neurological markers of depression is neuroimaging, predominantly magnetic resonance imaging (MRI). MRI has uncovered structural changes, functional alterations, or connectivity abnormalities in specific brain regions or networks in patients of depression. Thus, this project will focus on a retrospective analysis of the Canadian Biomarker Integration Network in Depression dataset, which provides inflammatory and MRI assessments in MLD patients. We aim to (1) synthesize a multimodal neuroimaging signature of MLD severity; and (2) assess the associations between this signature and systemic inflammation.

Year: 2024

Researcher:
Jean Chen, Baycrest

Student:
Muying Hui, University of British Columbia

Through SUDS, undergraduate students engage in hands-on research focused on data sciences and AI methodology applications.

Learn more

Oct 02 2026

Understanding the Neural Correlates of Attention in Children via Neuroimaging Data

Research description:

This project aims to develop a machine learning (ML) model for predicting children’s attention ability using features extracted from over 10,000 MRI images from the Adolescent Brain Cognitive Development (ABCD) study. The project will cover diverse data science topics, including big data in neuroimaging, network analysis, complex systems, feature selection, visualization, and cloud computing. The proposed features encompass those derived from T1 and T2 MRI scans, structural connectivity via diffusion, functional connectivity via resting and task-based fMRI, and non-linear metrics like fractal dimensions and Lyapunov exponents. The ABCD database provides predictive labels, including self-report surveys, clinical assessments (e.g., NIH Toolbox Flanker Inhibitory Control), and ADHD-related diagnosis and symptoms. Model interpretability is a priority. Feature selection should be transparent, and their respective contributions should be reportable and visualized. This project is part of a broader study on brain-computer interfaces and neural plasticity.

The student will have the opportunity to work with large-scale neuroimaging data, MRI/fMRI preprocessing, experimentation with feature selection methods and ML and/or deep learning models. They will have access to cloud computing resources and Google Vertex AI tools. The student will be supported by doctoral trainees and staff engineers in the lab. The expected deliverables will be:

A deep learning model trained on 10,000 MRI images from the ABCD study to predict children’s attention ability.
A feature visualization tool to qualitatively and/or quantitatively describe what brain areas and measures are physiologically relevant to attention in children.

Year: 2024

Researcher:
Tom Chau, Holland Bloorview Kids Rehabilitation Hospital

Student:
Melody Nguyen, University of Toronto

Through SUDS, undergraduate students engage in hands-on research focused on data sciences and AI methodology applications.

Learn more