Projets novateurs réalisés

Deep Learning Models for Principled Causal Forecasting

Numerous Machine Learning (ML) tasks are forecasting problems, used to make downstream decisions. Acting on ML forecasts however can changes the distribution of observations relevant to the forecast. The implication is downstream decision optimization procedures implicitly expect the ML model to generalize outside of the observational distribution. Unfortunately, this is often not the case, and ML models tend to be brittle outside of their training distribution. ML models will thus produce unreliable extrapolations, leading to poor downstream decisions based on wrong forecasts.

Causal models, which aim to learn the structural causal models underlying the data generation process, are a natural fit for such use-cases. This is because causal mechanisms are more robust to superficial changes in the data distribution, and can be expected to extrapolate better to new environments. This project aims to combine deep learning and causal inference to develop causal forecasting models adapted to two important applications. Students will start by implementing existing approaches on one of three applications, before working on improvements to core causal modeling techniques.

Hierarchical Connectivity Maintenance in Swarm Systems

The organization and control of swarm systems have been either completely decentralized where all the robots use only local observation and simple rules for coordination and cooperation which leads to emergent behavior. There are advantages to using this decentralized paradigm such as redundancy, scalability, and simplicity, but such systems tend to be slow and difficult to manage. On the other hand, using centralized systems in multi-robotic systems gives the ability to have a leader which makes managing the swarm easier and allows for faster task completion, but such systems tend to be difficult to scale and prone to failures. All in all, both methods’ disadvantages have hindered swarms to be widely applied in practice. Using a combination of these two points of view, we propose a hierarchical approach where the top-level agents are specialized in terms of hardware or compute power, and bottom-level agents are relatively simple. We plan to investigate this in the context of connectivity maintainance where agents in top level are intermittently connected and agents in the bottom level are continously connected.

The latent of Spectra fusion of X-Ray Fluorescence (XRF) and Mid-Infrared (MIR) Spectroscopy for detection and management of the Aluminium (Al) and Manganese (Mn) content in Canadian Soils for Potato (Solanum tuberosum L) cultivation.

The successful agronomic strategy relies on nutrient detection and estimation. Adequate management of nutrients is an important concern for the whole growth period of Potato (Solanum tuberosum L.) cultivation. In this work, we try to analyze the situation of potential harmful elements for potato production such as Aluminium (Al) and Manganese (Mn) in soil and plant, using spectroscopy, and find soil conditions that suppress their intake to the plants. based on soil attributes like pH, Nitrogen, Organic and Inorganic Carbon, water content and Cation exchange capability. The expected outcomes are :
i) To Detect and estimate Al and Mn concentrations in Soil and Potato plant
ii) To find the relationship between the situation of Al and Mn in the plant and soil and the effect of soil conditions on the uptake.
iii) To Develop prediction models of Al and Mn based on different machine learning techniques.

Conductive polymer deposition for textile strain sensors

Body movements are an important biomedical parameter that can be measured using wearable devices. Notable applications of wearable sensors range from monitoring individuals suffering from loss of autonomy to monitoring athletes’ performance. These movements can include breathing, speech, limb motion as well as heartbeat. Accurate measurement of these movements requires precise and sensitive strain and pressure sensors. In turn, the production of these sensors requires the development of reliable, robust, and highly conductive smart materials. Conductive polymers have been used to produce flexible and stretchable sensors and are popular in flexible electronics. They are therefore excellent candidates for conductive smart materials. By the end of the internship, the student will have developed an optimized method to polymerize monomers onto textiles for their use as electronic textile. Optimization of the method would lead to upscaled production of highly reproducible and reliable conductive threads usable for strain sensors.

Degradation of disposable face masks in the landfill leachate

The use of disposable face masks as preventive measures significantly increased in the past years. Overloading
landfills with disposable face mask wastes will raise certain environmental concerns. However, their degradation
in landfill leachate is poorly studied. The objective of this study is to investigate the degradation process of the
mask in landfill leachate and quantify the number of microparticles and chemical pollutants released into the landfill
leachate. This project will give the industry and government more detailed information about how to make more
improvements and act more Environmentally friendly in the future about the development and production of the
more degradable and biodegradable materials, considering the process of degradation.

Design and Development of a Mobile-based Medical Image Archiving System for Skin Cancer Screening

This project will help to design and develop the user interface and an image picture archiving and communication system (PACS) for mobile teledermatology for use in an application of skin imaging. This will enable both patients and specialists to acquire, archive and manage dermatological images for further diagnosis, triage and follow-up purposes. The project goals are to design, implement and evaluate the app interface of the new “MoleScope” dermoscope which attaches to a smartphone camera, in order to acquire and store images. One aspect of this interface is to provide a 2D body-map for users to locate the imaged lesions on the body for future follow-up. Also, this project will design and implement the desktop web-based interface for the specialist to evaluate acquired mole images (the clinical software). This requires implementing the communication module which will interact with a patient electronic management system, and includes

Decolonizing knowledge mobilization to advance Indigenous led conservation and reconciliation

The objective of this research project is to build capacity among Indigenous Nations and the
conservation sector to catalyze reconciliation through the establishment of Indigenous Protected and
Conserved Areas (IPCAs). To do so, this we will investigate how knowledge and evidence-based
research can be effectively communicated in cross-cultural and Indigenous contexts. While the field of
“knowledge mobilization” offers researchers a wealth of information on turning research into action, the
field has been slow to adopt decolonial approaches. Partnering with two Indigenous-led
organization/initiatives—the IISAAK OLAM Foundation and Conservation through Reconciliation
Partnership—and informed by many collaborators, we will generate a diverse suite of informative
outputs. In addition to open-access articles, these may include guides, educational modules, summaries,
infographics, evaluation frameworks, etc. This project will directly support the IISAAK OLAM
Foundation and Conservation through Reconciliation Partnership, including two major initiatives: IPCA
Knowledge Basket and the Pacific IPCA Innovation Centre.

Examining social innovation engagement models to promote social cohesion and equity via climate action

This project looks at the drivers, impact and actor-groups that lead to successful Social Innovations.
Social Innovation has prominence in both the public and private sectors, as a way to reform societal relations and create system level change, in the structure of relationships with the goal of empowering community actors to lead the development of solutions to challenges impacting them.
There is a general consensus that SI has positive results; however, there are questions in academia around whether SI actually results in improvements.
The work of Inspiring Communities will be observed and analysed to deeply understand the organization’s social innovation approach and reveal the impact of processes on systems level change. Insight into how social cohesion impacts the effectiveness of initiatives will inform a case study on how a rallying point such as climate action can be used to engage communities and sustain resident participation in interventions in novel ways

Analysis and optimization of conceptual landing gear designs

This project, sponsored by Safran Landing Systems Canada (SafranLS), involves the development of engineering techniques that enable rapid conceptual design of aircraft landing gear. The goal is to speed the early design of new landing gear to enable SafranLS to produce more, faster bids to manufacture landing gear for new aircraft at their two Canadian sites. The interns engaged on this project will be using existing SafranLS data to improve models for kinematic optimization and parameter estimation, and implement these models in computational codes that are intended for use by SafranLS engineering staff. In total, there will be eight four-month internship units among six interns during the course of the project. These interns will get critical training in mechanics and design of landing gear, and will interact with the SafranLS design engineers throughout their internship terms.

Field Assessment of Infrastructure Carrying Capacity

Over the last five years, it has become possible to estimate the capacity of existing, sometimes damaged,
utilities and other infrastructure systems to support the local population in war zones. This estimation
currently relies on satellite data to provide a basic map of what infrastructure is where. However, it is
often difficult o get access to current satellite data while the war continues, amid security concerns by
either side. Therefore, a simpler approach is needed that allows relief and reconstruction planners to gain
sufficient understanding of the actual infrastructure capacity using only field observations and interviews
by local Red Cross personnel. This research seeks to establish generic models for each critical
infrastructure system to inform that field data collection. Success would represent a significant step
towards more efficient and timely reconstruction planning that directly benefits the local population and
alleviates suffering by the most vulnerable.

Investigating astrocyte responses in brain metastasis of HER2-positive breast cancer

Brain metastasis of breast cancer has a very poor prognosis in patients. Treatment options are limited because the blood-brain-barrier prevents drugs from entering the brain. Importantly, we are lacking information on how brain cells are supporting the growth of metastatic breast cancer cells. Using HER2-positive breast cancer cells isolated from a patient brain metastasis the Hombach lab has established a mouse model to study the responses of astrocytes, the most abundant glial cells in the brain, to brain metastatic tumors. Astrocytes will be isolated from the mouse brain at early and late time points of metastatic development to determine changes in their gene expression profile. We will validate selected genes and will use mouse brain tissue sections to study their localization. This will allow us to identify the responses of astrocyte in the vicinity of brain metastasis and inform us on novel therapeutic targets to explore in the future.

Les as de l’info : analyse d’un dispositif numérique devant contribuer à combattre la désinformation chez les jeunes et leurs familles

Dans le contexte communicationnel et technologique actuel, les enfants se retrouvent face à une prolifération d’informations et de contenus. Cette dynamique ouvre grande la porte un phénomène de désinformation poussé, tout particulièrement, par la communication numérique. Or, le numérique est aussi porteur d’opportunités de contre-information et les enfants ont la capacité de se montrer critiques et actifs face aux contenus qu’ils consomment. Dans le cadre de ce projet de recherche partenariale mené avec le média jeunesse numérique Les as de l’info (média destiné aux 8-12 ans), nous désirons évaluer le rôle que peut jouer ce dispositif en ligne dans le combat contre la désinformation chez les jeunes et leurs familles.