Projets novateurs réalisés

Boosting Robustness of Deep Neural Networks against Sparsity-aware Adversarial Attacks

Over the past few years, deep neural networks (DNNs) have been used to solve a wide range of real-life problems. However, DNNs are vulnerable to adversarial attacks where carefully crafted input perturbations can mislead a well-trained DNN to produce false results. As DNNs are being deployed into security-sensitive applications such as autonomous driving, adversarial attacks may lead to catastrophic consequences. In this research project, we focus on those adversarial attacks that target energy-efficiency of DNNs. In particular, there are a group of attacks that insert perturbation into inputs of neural networks so that the number of zero values reduces. This will increase energy consumption and latency of machine learning applications. We propose a solution to monitor outputs of neurons and detect these malicious attacks dynamically and in runtime. This project enables TrojAI to offer machine learning algorithms that are protected against sparsity-aware adversarial attacks.

Systèmes sages et DIY (Do It Yourself) au service de la santé

Le projet Mitacs Globalink que nous proposons est la première étape d’une collaboration entre le laboratoire DOMUS de l’Université de Sherbrooke et le laboratoire LISTIC de l’Université Savoie Mont Blanc. Dans sa thèse, M. Hubert Kenfack NGANKAM a créé un système de création de scénarios pour aider les personnes atteintes de la maladie d’Alzheimer lors de l’errance nocturne, mais les soignant.es sont toujours impliqués dans le processus de création de ces scénarios. Ce que nous avons proposé est de combiner notre système qui peut se connaître par lui-même appelé système sage avec le système de création de scénarios pour alléger la charge de travail des soignant.es, de plus, nous pouvons aller plus loin et augmenter l’application nommée “Virtual Adviser” créée par l’équipe DOMUS et installée sur le casque AR Microsoft Hololens en proposant des appareils connectés que les soignant.es peuvent acheter lors de la sélection d’un scénario.

Understanding Indo Canadian Female Challenges After Covid-19

Many skilled women workers are going through unemployment phase after Covid-19. There few are under employment. These skilled, and educated professionals are facing domestic work pressure, and economic challenges, where the secauses unhappiness in their life. This Covid-19have impacted badly to the life of IndoCanadians, particularly women as a vulnerable group. However, this research addresses the questions as,1.How do Indo Canadian women perceive their identity after Covid-19? 2.Whatare the Indo Canadian women’s integration challenges (violence, unsafe environment, discrimination and so on) at workplace? Canadian development agenda opens a platform to discuss,and debate about the integration challenges to provide equal opportunities, and gender based equal treatment for IndoCanadian women. To put in a nutshell, in Canada, a current social, and economic challenges relational impact due to Covid-19. Indo Canadian women are struggling for a better life, empowerment, and emancipation

Evaluating diverse non-brewing yeast strains for novel new-make spirit applications

Many of the flavours found in whisky and other distilled spirits are derived from alcoholic fermentation, most commonly by the yeast Saccharomyces cerevisiae. Given the biodiversity of different yeast strains and species, there is an opportunity to identify strains that can produce alcoholic beverages with unique flavours and aromas. The Measday lab has isolated hundreds of yeast strains from uninoculated fermentations of wine grapes from the Okanagan Valley, BC. Most of these strains can grow on maltose as a sugar source which suggests the ability to ferment wort, the precursor for beer and many spirit fermentations. Our research proposal will validate which yeast strains have traits suitable for commercial use in whiskey production. The results of this project will benefit Driftwood Brewing Company’s distilled beverage design and quality control and exemplify the benefits of using diverse BC yeast for spirit distilling applications.

Moving home: Improving reintroduction outcomes and human-capuchin monkey coexistence in Brazilian caatinga and forest fragments.

In this project, we bring together a team of four proposed MITACs globalink student scholars working in Brazil and Canada with supervisors directly involved with the research project. We will begin to fill the gaps in knowledge about successful reintroduction of capuchin monkeys by 1) conducting behavioural research on bearded brown capuchin (*Sapajus libidinosus*) personality variables among pre-release captive monkeys during rehabilitation; 2) collecting data in the field post-release to measure habitat use through observation, trail cameras, and GPS collar tracking, and 3) analyzing data on direct field follows of habituated wild blond capuchins (*Sapajus flavius*) living in a forest fragment to examine group movement patterns and habitat use. The results from this research will improve human-primate coexistence and biodiversity conservation management, both directly in Brazil and indirectly for similar wildlife and behavioural ecology projects in Canada and globally.

L’accompagnement entrepreneurial : une approche relationnelle et écosystémique

Les entrepreneurs et entrepreneuses ont souvent des parcours hétérogènes, peu conventionnels et pavés de détours. Or pour trouver un accompagnement adapté aux étapes de la mise en place de leur entreprise, plusieurs se tournent vers l’accompagnement entrepreneurial. Celui-ci peut être à la fois du mentorat, de l’apprentissage par les pairs, mais également à travers des structures d’accompagnement, des incubateurs, des accélérateurs. Espace inc., un accélérateur entrepreneurial, déploie présentement un modèle suprarégional d’accompagnement tout en proposant une offre de formation par les pairs.

Ce projet se propose donc de documenter et d’analyser les pratiques d’excellence en accompagnement entrepreneurial. En documentant et analysant (par des entrevues, de l’observation, de l’analyse de données internes) la relation qui se développe entre les entrepreneurs et leur pair-enseignant de même que les défis et enjeux entourant l’accompagnement suprarégional, nous espérons proposer un modèle qui tienne compte à la fois des dimensions interpersonnelles et écosystémique régionale de l’accompagnement entrepreneurial.

Medzy : Pharmacien Virtuel – Triage et aiguillage intelligent des patients

Medzy est une plateforme de pharmacie virtuelle qui assure un suivi thérapeutique assidu, fondé sur un système de rappels et d’alertes, un dosage de médicaments clair et personnalisé, ainsi qu’une interaction continue et entièrement confidentielle entre le patient et son pharmacien. Ce dernier devient donc un acteur proactif du cheminement thérapeutique. Le patient, lui, voit sa vie grandement simplifiée et profite d’une expertise soutenue.
Medzy a pour projet de créer une passerelle d’information qui permettra aux parties prenantes en lien avec la santé du patient, notamment le médecin et le pharmacien, d’assurer, ensemble, un meilleur suivi thérapeutique global et ainsi diminuer la charge sur le système de santé.

Digital Pre-Distortion for Concurrent Multiband and Multiple Antenna Transmission

The proposed project aims at the design and implementation of low complexity digital pre-distortion (DPD) algorithms for multiband and multiple input multiple output (MIMO) wireless transmitters. The power amplifier (PA) is one of the major sources of power dissipation in wireless base stations. The DPD techniques enable the PA to operate in a more efficient power level resulting in more energy efficient wireless networks. To address the ever increasing demand for higher data rates, the MIMO and the carrier aggregation techniques are being adopted by the current and upcoming wireless standards. The requirement for concurrent transmission over multiple bands in carrier aggregation schemes and over multiple antennas in MIMO schemes poses a great challenge to the design of feasible DPD algorithms. The multiband/MIMO DPD algorithms currently available in the literature are computationally too complex to be implemented in practice. TEKTELIC believes that the proposed research project will result in low complexity and yet effective DPD designs that can be used in the company’s future multiband and MIMO radio products.

Pipe replacement and rehabilitation optimization method

The research project will develop a pipe replacement and rehabilitation optimization method. The research group is currently developing machine learning models to predict future water main failure. Data has been collected from 13 cities across Canada and represent different conditions and settings. This data is already being cleaned and analysed by a team of graduate students. This project will focus on one specific city as a case study to develop an optimization method. This method will build upon the predictive models being developed and will establish a risk framework for replacement and rehabilitation prioritization, accounting for economic, social and environmental impacts.

Solid Waste Management Best Practices: Cost-effective options to sustainably manage solid waste in the Peace River Regional District

Concerns over climate change have led public and private institutions to change the way they conduct business. Initiatives include ways to reduce greenhouse gas emissions and working toward a net zero carbon society. Recycling and composting programs have been around for decades, with the goal to reduce the amount of solid waste entering the landfills. The natural decomposition of food and green wastes in landfills results in the emission of methane, a potent greenhouse gas. Plastics in the environment find their way into the food chain and even human tissues. Solid waste (SW) diversion is a strategy that reduces landfilling and greenhouse gases, and recovers energy and materials.
The PRRD is located in northeast BC, a distance of 800+ km to major recycling centers. The PRRD is seeking to implement a SW diversion strategy that sees recyclables processed in the region, avoiding long-range transport, with a novel waste to energy plant. The environmental impact must be assessed to determine the economic and environmental feasibility. An economic evaluation has already been performed.
For food wastes, anaerobic digestion (AD) has been selected. The process captures >95% of methane emissions.

Representation Learning in Knowledge Graphs

Mastercard is looking to make transactions simple and secure for its customers and vendors across the globe by harnessing the power of graph learning. In this project, the power of graph-based deep learning will be utilized to develop a system that can detect fraud entities, anomalous communities, etc. This track will enable card users and online buyers and sellers to enjoy safe and secure digital transactions. The final goal of knowledge representation learning (KRL) from graphs is to obtain a universal information rich embedding that can be used under multiple applications. The benefit to the partner organization includes a greater understanding of the domain of Knowledge Graphs, identification of gaps in the current area of research, and the possession of a system that can detect fraud entities with better accuracy compared to the state-of-the-art systems in the field.

Predicting Perfect Matches Between Organizations

Successful partnerships between non-profit organizations and funders are fundamental for organizations’ abilities to concentrate on their mission, and for founders to see their philanthropic investments align with their values and make real impacts. Matching founders and organizations has traditionally been done informally through personal connections, networks, and expensive prospecting software and databases. Just like a dating app for philanthropic organizations, Orgmatch matches funders and non-profits together based on attribute-based representations. This research project will explore the use of Artificial Intelligence/Machine Learning technologies to perform this matching Non-profit organizations are fundamentals to a healthy society and the philanthropic relationships with the funders sustaining them are essential to their survival. A tool to rapidly and successfully match them would allow both parties to focus on results and real impacts to the community.