Innovative Projects Realized

Forecasting Patient Flow Pressures

At St. Michael’s Hospital (SMH), having insight into patient flow throughout the hospital is essential to resource planning and operational efficiency. When patients are admitted, discharged, or transferred in the hospital, several actions need to be taken to ensure patients receive timely care and resources do not become backlogged. Improving patient flow reduces wait times, improves operational efficiency, and ultimately improves care. However, due to reasons like the epidemic outbreak, traditional methods of controlling the flow of patients are no longer effective. In this project, Unity Health Toronto is seeking to find better way to control the patient flow, efficient by utilizing state-of-the-art machine learning models.

Fabrication of Deformable Micro Mirrors

Histology studies the microscopic anatomy of biological tissues, which is critical to clinical disease management and fundamental to biological understanding. However, current approaches that rely on bright-field microscopy require extensive issue preparation prior to imaging and may thus take days. Photoacoustic remote sensing, or PARS, can dramatically reduce the histology bio-imaging time from days to minutes. PARS microscopy must be capable of 3D imaging or optical sectioning, which enables visualizing multiple layers of disease tissue without the need for lengthy physical sectioning. This in turn calls for auto-focusing or adaptive optics, to focus on different depths of the tissue (i.e. optical sectioning). In this project, we will fabricate resonant deformable micro-mirrors that can shift the focus by ~10cm as needed for high throughput PARS microscopy, with mirror actuation voltage down to 20V. Besides auto-focusing, the mirror can also correct aberrations of the optical system. Moreover, the resonant micro-mirror can avoid the high electrode count (on the order of hundreds), complex drive electronics and control software, as needed by the conventional DC/static actuation of deformable micro-mirrors.

Modeling Application Performance under Multi-Instance (Multi-Stream) Execution Scenarios

Multi-stream execution is a technique in GPUs that allows multiple operations/kernels from the same program to effectively use GPUs without explicitly stating the affinity of threads to the cores. Several recent optimizations in Machine Learning (ML) algorithms leverage multi-stream execution. While performance modeling of ML applications is well studied under single-stream execution, performance models of novel ML applications under multi-stream execution is lacking. There is a pressing need to develop performance models for multi-stream execution – that would be the primary area of exploration under this co-op/internship. Specifically, we are expecting to study the state of the art (SOTA) in performance modeling for multi-stream execution and develop first principles performance models, conduct validation with silicon performance, and integrate the performance models in an internal simulator developed at AMD.

Optimized operational strategy planning model for Smart Net Zero Community

Due to some complex technical problems, both power plant’s generators and local grid transmission lines do not have the potential to generate and transmit enough electricity to cover and meet all load demand during peak hours. Based on this technical difficulty, there is no solution for customers to use costly energy during peak hours and accordingly pay more money for their consumed electricity(Refer to electrical pricing scheme).In this case, in order to remedy above mentioned problem, smart solar communities as grid-friendly consumers, would be established with the purpose of supporting the local grids. This feature would be more important when local grids are in peak period hours. To this end, solar communities should shift their internal electric loads from peak load period to off-peak hours. Regardless of shifting the loads, they should either decrease overall amount of energy consumption or used dynamic load shedding/shaving methods. In this case, for managing load demand during peak hours, a community-scale optimal strategy planning model is necessary. Base on community-scale dynamic complexity and uncertainty, this model should be constructed based on inexact optimization techniques such as Interval Parameter linear Programming, (Mixed) Integer Linear Programming, Fuzzy Linear Programming or integration of above methods.

How Wildlife Use Existing Crossing Structures along Roads in the Laurentians: Light at the End of the Tunnel?

Roads and traffic subdivide wildlife habitats and increase the mortality of wild animals that try to cross due to collisions with vehicles. Wildlife crossing structures are costly, but an alternative may exist in structures that are already present under roads such as water culverts. It is unknown how often wildlife use such structures in the Laurentians. This project will install cameras on many of these structures and track-boxes near them to identify the species that use the structures and quantify their frequency of use. Statistical analysis of the data will help Éco-corridors laurentiens evaluate what future measures should be put in place to further reduce the negative impacts of roads on wildlife. This is very important in the Laurentians, as the road and highway being evaluated pose a threat to wildlife frequenting National Parks in the region.

Design of Tunable Surface Coatings for Magnetic Nanoparticles for the Adsorption and Removal of Organic Contaminants

This project entails the design, synthesis, and characterization of porous surface coatings made from metal–organic frameworks for magnetic nanoparticles. The combination of these two materials will allow for the removal of organic contaminants that currently interfere with the correct functioning of DNA and RNA extraction test kits (e.g., COVID-19 test kits). As developers and producers of such kits, the partner organization (Galenvs Sciences Inc) will benefit from this novel joint research approach by improving the quality and function of these kits. In addition, Galenvs will enhance the size of their materials research portfolio for incorporation into future products.

Efficient, Edge-based Video Streaming for Video Action Recognition

Streaming video applications in the home (video doorbells, security cameras, babycams, robot vacuums etc.) have started to become commonplace. Automatically detecting events of interest in such video streams can help to reduce manual effort from users in sifting through false alarms, enhance user safety, comfort and satisfaction, as well as enable new applications in areas such as safety and wellness. This project focuses on systematically investigating the performance of action recognition models under resource constraints found in practical deployment scenarios such as variable and metered network bandwidth, limited compute, etc. as well as develop technologies to overcome these challenges.

Équité algorithmique; impacts et changements de mesures

Présentement, les modèles de tarification en assurance auto calculent les primes chargées aux assurés en tentant de prédire leur risque d’assurance grâce à plusieurs critères (âge, sexe, état-civil, expérience de conduite, caractéristiques du véhicule, etc.). Le problème auquel ce projet s’adresse est celui du point de vue éthique d’une telle approche. En effet, l’usage de certaines variables peut être vu comme discriminatoire envers certains groupes de la société. Un des exemples courants est que les femmes présentent généralement un risque d’assurance moins élevé et paient ainsi des primes d’assurance plus basses. Cette pratique est généralement acceptée socialement et permise par la législation canadienne, mais il pourrait y avoir d’autres situations similaires concernant d’autres variables sensibles. Le but général du projet sera de déceler ces situations, de déterminer comment les améliorer si nécessaire et d’évaluer l’impact potentiel d’antisélection que les modifications auront au global sur les portefeuilles d’assurances concernés.

Ceci est avantageux pour l’organisme partenaire dans la mesure où il souhaite s’assurer que ses pratiques de tarification soient alignés aux valeurs de l’entreprise.

Predicting Equipment Failure in Canadian Mines

In order to reach the production targets, mining operations typically need to operate without a stop. Equipment maintenance planning plays an important role in ensuring the required number of mining haul trucks are available at any given time. If not planned correctly, corrective maintenance would be required, which could slow down or halt the production as well as potentially costing higher in maintenance. In this project, reliability analysis and predictive maintenance for turbocharger and transmission failures will be proposed using physics-based, data-based, machine learning based and hybrid approaches, using the data collected by the sensors on mining haul trucks. The factors that increase the fuel consumption will also be studied. Finally, anomaly detection model will be developed to detect sudden sensor reading changes and failures for the cases to alert the driver or an operator.

Variations du microbiome des baies et des feuilles de vignes soumises à différents changements de température au champ

Au Canada, les études explorant la relation entre la vigne et son microbiome sont encore peu nombreuses, surtout
sur les cépages hybrides cultivés dans l’Est du Canada. Ces cépages, issus de croisements entre l’espèce Vitis vinifera et des espèces Vitis indigènes à l’Amérique du Nord (V. labrusca, V. riparia), ont une biochimie particulière, qui leur permet une forte résistance au froid et une maturité hâtive, mais qui affecte aussi la biochimie des baies.1,15
Les modèles de changements climatiques suggèrent que l’Est du Canada connaîtra des augmentations significatives des températures dans un avenir proche.16,17 Ces augmentations auront un impact global sur l’agriculture canadienne, notamment dans le secteur de la viticulture. Si l’on peut s’attendre à un développement plus hâtif et/ou plus abouti des baies, la température affectera également la microflore des vignobles, avec un impact certain sur l’incidence des maladies, mais aussi sur la qualité des vins produits. Dans le cadre de ce projet, nous caractériserons le microbiome fongique et bactérien du cépage hybride l’Acadie blanc cultivé dans l’Est du Canada puis nous évaluerons sa résilience face à différents traitements de température mimant les changements climatiques.

Portfolio Optimizer

Portfolio managers evaluate stocks to find those with the best expected returns, but they need risk measures to complete their decision-making process.
The project has the objective to give adequate risk metrics to portfolio managers for a given portfolio and a list of potential candidates.

Historical price values are useful to determine risk, but they could be better optimized to reflect the uncertainty given that historical values may not reflect future values.

The model which uses reinforcement learning should give portfolio managers better insights on their risk when they add a new stock to their portfolio.

The goal of the partner organization is to improve its knowledge on the matter as well as offering free access to the research made by the intern.

Expanding the analytical capacity of Nanopore sequencing for infectious disease surveillance and diagnostics in a One Health framework

The alarming increase in the scale at which infectious diseases spread and devastate global socioeconomics necessitates timely characterization of the disease-causing agent to devise effective intervention strategies to stop the spread of disease at the source of introduction. It is known that domestic and wild animals play significant roles in the transmission of pathogens to humans via direct contact or the ingestion of contaminated foods. Hence, the proposed project will partner with an animal health diagnostic lab, Prairie Diagnostic Services Inc. (PDS), to develop laboratory and computational solutions to expedite decision making in clinical and surveillance settings. The tools and protocols will replace laborious conventional workflows with long turnaround times and streamline the process from pathogen sequencing data generation to result interpretation. The generated genetic information of the pathogens from the project will subsequently be used to identify genetic changes associated with increased livestock mortality rates and disease prevalence. The collective work will provide PDS with the analytical infrastructure to offer accredited rapid sequencing services for Canadian researchers and agriculture whom will benefit from the assurance of food safety and livestock productivity.