Innovative Projects Realized

Out of Distribution Detection in Deep Generative Models

As generative models become increasingly prominent in machine learning, the need for accurately detecting out-of-distribution data has become crucial. The primary objective of this research is to develop an approach that can identify when the program encounters data that is vastly different from what it was trained on. In machine learning, programs may make errors when they encounter data that is dissimilar to what they have learned. To tackle this issue, we will investigate various techniques utilizing deep generative models to help the program comprehend what types of data it should expect to encounter. However, even the most sophisticated deep generative models may occasionally mistake new data as similar to old data, leading to inaccurate predictions. Therefore, we aim to investigate the underlying reasons for this phenomenon and explore potential solutions to address this issue.

A putative chloroplast Ustilago maydis effector causes morphological changes in Arabidopsis thaliana

Plants have several ways of defending themselves from plant pathogens including physical structures such as thick cuticles and defense hormones such as salicylic acid. The latter induces a cascade of plant defense responses that can ultimately lead to resistance. Similarly, to successfully invade their host, plant pathogens secrete a cocktail of proteins called effectors that favour the virulence of the pathogen. We identified seven candidate effector proteins from the corn smut fungus, Ustilago maydis, that potentially target the maize host’s chloroplasts, which are the organelles where the biosynthesis of salicylic acid happens. The overexpression of these seven potential effectors in the non-host Arabidopsis thaliana led to the foundation of this project. When overexpressed in A. thaliana, I found that one potential chloroplast effector caused a substantial morphological change in the plant, despite no changes in its susceptibility to the bacterial pathogen Pseudomonas syringae pv, maculicola. I plan to use the exchange in Germany to identify the plant interacting partner of this effector protein. This will further our understanding of the role of different effector proteins in causing morphological changes in their host in order to facilitate their development in planta.

Exploring an interactive multisensory physical movements model during and beyond COVID-19: a case study of children with special needs

The COVID-19 pandemic has forced children to quickly adapt to home-based or virtual learning; however, a number of researchers have identified challenges and difficulties with applying and using technology. More importantly, there has been a significant rise in the rates of mental illness occurring as a result of COVID-19 and children are now experiencing increased mental health and physical challenges. This study is to conduct ongoing interdisciplinary research which aims to assist children by adopting a novel multisensory model during and beyond COVID-19. The primary expected outcome is to provide an innovative method combining different senses and emerging technologies to improve children’s attention and enhance teaching effectiveness. Qualitative and quantitative data will be collected to assist practitioners and parents who currently have no related reference to serve in the process of creating evidence-based guidelines. The research findings will broaden the scope of current literature focusing on the need to emphasize the importance of empirical investigations into how best to use the multisensory model within the learning environment.

Arctic Research Foundation– Polar Data Analyst Project

The Arctic Research Foundation (ARF) is a private, non-profit organization creating a new kind of scientific infrastructure for the Canadian Arctic, through its operation of efficient, cutting-edge research vessels and self-powered mobile labs. This phase will concentrate on extending and improving the functionality of the platform, and will include developing processes to mint DOIs and a means for researchers to request large file uploads. In addition, a framework for developing new types of visualization will be defined and implemented along with UX improvements to the admin area of the platform.

Development of Data Collection Procedures

1JustCity’s West End Drop-in provides support to West End community members seven days a week. The main part of their organization is a drop-in lunch program, but there is a constant push for their organization to do more than that. They provide Indigenous Cultural Programming, offer Housing Support, and aim to create a community where people can form an identity outside of poverty. 1JustCity is growing quickly, and although grateful for opportunities to increase funding, there is an increased expectation of tracking data that has been a challenge to keep up with. This project will help the organization develop a concrete data collection process, allowing 1JustCity in the West End to track and save information about daily visitors. The project will work to improve the internal processes of data collection within the organization by having a more efficient and effective way to track data. This will allow the organization to increase their sources of funding.

Sales Process for medical technology

Foqus has built a software solution that can speed up MRI scans and is applying for regulatory clearance to commercialize the software. This project focuses on the sales process of the product and facilitates taking the software to the market.

Brett McGonigal – Studying Advanced Manufacturing Strategies for Autoinjectors

The intern will participate in an interdisciplinary project on advanced manufacturing strategies for autoinjectors, which have become increasingly popular due to their convenience and reliability in administering medications. While the Epipen is a well-known example, other medications, such as Noxalone, use this delivery system.

The scale and extent of autoinjector manufacturing in Canada are unclear, and further research is required to determine the current status. One significant obstacle with the current manufacturing process is the high cost associated with producing autoinjectors, primarily attributed to material, assembly, quality control, and regulatory compliance costs. Identifying the specific parts of the manufacturing process contributing to the high cost is crucial in finding ways to overcome them.

The intern will collaborate closely with the Cansbridge Fellowship, the partner organization, and their academic supervisor, Dr. Alan Ableson, to investigate the latest technological advancements in advanced manufacturing in Canada to optimize cost efficiency. For example, the project will explore whether automation can streamline the assembly process or whether 3D printing can reduce production costs by removing the need for expensive molds.

The intern will benefit from the Cansbridge Fellowship’s extensive alumni network, which includes leaders from top innovation-driven organizations like Google, Tesla, and numerous startups.

Paul Yang — AI/ML Data Analysis of Fashion Consumer Trends in Social Media: An Impact Report for Canadian Businesses

The project, “AI/ML Data Analysis of Fashion Consumer Trends in Social Media: An Impact Report for Canadian Businesses,” aims to analyze fashion consumer trends on social media platforms and produce an impactful report for Canadian businesses. The intern will employ and develop advanced data analysis techniques to collect and process data from various social media sources to gain insights into the latest fashion trends, which will be presented visually in an analytical report. This project is of great importance to Canadian businesses as it will provide a comprehensive overview of what consumers are looking for, enabling companies to improve their products and services accordingly, leading to an improvement in the economy. The intern will work in collaboration with the partner organization, Cansbridge Fellowship, and academic supervisor, Alan Ableson. The project’s direct benefit to the partner organization is the production of a real-world case study, which can be used to assist other project’s within the Cansbridge Fellowship program as a whole. This will further attract Canada’s best leaders in entrepreneurship and leadership.

Évaluation de modèles neuronaux pour la correction de textes en anglais et français

THIS IS A GENERIC TEXT PUT IN PLACE AS THERE WAS NO PROJECT OVERVIEW

Fusion d’une méthode classique et d’une méthode par apprentissage automatique pour la détection de mouvements humains par radar

Le radar développé par Morphée+ détecte avec une grande précision la présence de 1 à 3 sujets/objets mobiles dans une pièce vide ou avec des objets immobiles. Cette identification des mouvements est affectée lorsque d’autres objets ou sujets sont en mouvement dans la pièce. La méthode utilisée actuellement par Morphée+ se base sur l’utilisation de modèles mathématiques et statistiques qui s’appliquent au fonctionnement des radars mais aucune solution satisfaisante n’a été développée à ce jour.
Un premier projet de recherche en amont a pour but d’identifier des modèles d’apprentissage automatique pour traiter le signal radar et séparer le mouvement d’une personne des sources de bruits.
Ce projet-ci vise, quant à lui, à combiner la méthode mathématique utilisée actuellement par Morphée+ et les méthodes basées sur l’apprentissage automatique qui auront été identifiées lors des travaux de recherche précédents.

Enhanced machine learning for hematology analytics with near-field microscopy

Alentic Microscience Inc has developed a highly portable device that can analyse blood in a few minutes, opening the possibility for much improved healthcare for a lower coast. The system uses AI a proprietary sensor system that provides data for an AI based analysis technique. In this project we will evaluate the impact of recent developments in attention-based AI systems, which is a principle ingredient in systems like ChatGPT. These systems can potentially enhance considerably the capabilities of the blood monitor.

Automated Demand Response Considering Consumer Satisfaction and Congestion Relief Potential

This study develops a smart building HVAC management system considering consumer’s and distributor’s perspectives. A mathematical model will be formulated such that the total cost and dissatisfaction are minimized and the building’s potential to provide congestion relief in the area will be measured. In order to overcome the complexity of the problem, a group of deep reinforcement learning methods will be developed to manage the electricity flow in the building. The results will be validated by the real data provided by BrainBox AI and will be applied on some different clusters of the consumers. Moreover, some cost and benefit investment analysis will be carried out to provide policy making recommendations.