Projets novateurs réalisés

Metamorphic Testing of Deep Neural Networks

Nowadays, we are witnessing a wide adoption of Machine learning (ML) models in many software systems. They are even being tested in safety-critical systems, thanks to recent breakthroughs in deep learning and reinforcement learning. Bombardier, a world leader company in the domain of aerospace have started integrating deep learning models in its products, systems and processes design used for certification and homologation. Although these ML models can reach high accuracy within their training data distributions, recent studies have shown that they can be very brittle. Hence, they should be tested adequately. However, because of the absence of a reference oracle, it is very challenging to reason about the behavior of a Deep Neural Networks (DNNs) and hence to test it. Novel testing techniques are needed both during model engineering and deployment phases, to guarantee the reliability and robustness of in-production DNN-based software. During the model engineering phase, developers need to assess the impact of their configuration choices carefully. The effectiveness of this assessment depends on the capability of testing data to trigger both the major functionalities of the model (regular cases) and the minor functionalities (corner cases).

Amélioration des protocoles de cryoconservation des semences de taureau en utilisant des protéines de blé comme agents de cryoconservation

L’industrie laitière et de bétail implique la distribution de semences génétiquement optimales et de qualité. La cryoconservation du sperme bovin est la façon la plus efficace de fournir ces semences aux producteurs à travers le monde. L’inconvénient est que ce processus de congélation, à de très basses températures, induit une diminution de la fertilité des spermatozoïdes. Les industries fournissant le sperme doivent alors améliorer leurs protocoles de congélation afin de répondre à la forte demande des producteurs. De nombreux protocoles ont été établis, cependant les résultats ne sont pas assez satisfaisants ou montrent des contraintes telles que de la toxicité. L’objectif de Boviteq est de développer un nouveau protocole de cryoconservation du sperme impliquant des protéines de blé et de comprendre les mécanismes de protection des spermatozoïdes par ces protéines afin d’optimiser les protocoles.

Social impacts on users and community : methodological approach and a sustainable building case study

The buildings have an important environmental and social footprint, and to respond to that, the sustainable buildings are the new tendency of the building sector. To reduce the negative impacts associated to it, accreditations like LEED were developed, but it is mostly focused on the environmental aspects, and the social criteria are sadly neglected. This project research aims to develop a methodological approach to measure the social impacts of sustainable building projects, on the users (residents, employees) and the local community. The methodology includes a literature review on various field of research (architecture, urbanism, environmental psychology, social life cycle assessment), a series of semi-directed interviews with experts in these fields, and finally a case study of a sustainable building. The internship at Gestion Immobilière Quo Vadis will enable to develop the appropriate tool to evaluate the social performance of a sustainable building.

Effet des structures modulaires reconfigurables et des cellules dynamiques sur la performance des PME manufacturières au Québec

De plus en plus, l’amélioration des processus de fabrication et des méthodes de travail est un enjeu au cœur des préoccupations des petites et moyennes entreprises au Québec, et ce, encore plus dans le milieu agroalimentaire en raison de la diminution notable du nombre de petites fermes. L’incertitude et les variations constantes de la demande dans ce secteur forcent les entreprises à revoir leur productivité, aller vers l’international et revoir leur façon de faire pour être le plus flexible possible. L’utilisation de cellule dynamique peut donc être une voie extrêmement importante à prendre pour des entreprises souhaitant se démarquer par sa flexibilité de production. En effet, ce principe permet d’optimiser la production d’entreprises souhaitant se diriger vers un marché de commande par catalogue.

Toward Building a Technical Engineering Smart Search

Documentation is an integral part of any product development. Product and documentation should co-evolve; whenever the product is changed the documentation should be updated and vice versa. Yet, oftentimes, multiple documentation and product versions may exist; between and within the documentation of the product, relations and dependencies may or may not be explicit, documented, trustworthy or consistent. For example, dependencies may not be explicitly stated (or different terms may be used for the same concepts), obscure jargon and technical abbreviations may be used inconsistently. In a regulated domain such as aerospace, a lack of consistency and–or coherence in the documentation of a product is a critical issue. Airborne software certification requires documentation to be unambiguous, precise and complete. Documentation and documentation relations are vital to ensure and prove dependability.
The high-level and long-term goal of this research is to support engineers in information gathering and fruition via semi-automatic document analytics (i) to promote information consistency; (ii) to extract and build trustworthy document (and document parts) relations (including traceability relations) and models; and (iii) to foster an effective information access and use.

Impact of Wind Energy Turbines on Bat and Bird Mortality

In this MITACS-funded project, a postdoctoral researcher will work with partners at Trent University and Bird Studies Canada to expand our understanding of how wind turbines affect birds and bats. We will leverage an extremely detailed database on wind-wildlife interactions that is managed by Bird Canada. Using these data, we will investigate whether bat and bird mortality are affected by turbine characteristics including height and the area swept by the turbine blades. This work directly supports Birds Canada’s mandate to conservation wild bird populations, and our results will contribute to developing renewable energy without compromising biodiversity.

Intelligent Chatbot Development for the Tourism Industry

This research helps to automate ontology development in order to support semi-supervised and active learning chatbot as much as possible, so that the overhead of chatbot training that requires human supervision is minimized, while relevant knowledge management activities become more efficient. The research objectives are both to refine the quality of the chatbot interactions and to automate its development and training as much as possible, to implement and test its practical and cost saving capabilities in tourism industry.

Machine Learning for Breath-Based Cancer Diagnosis

Non-invasive breath analysis has substantial potential for monitoring of a wide range of medical conditions and observation of overall health status. Breath testing is easy and painless; it can be done quickly and inexpensively, and can be repeated as often as needed, making it an attractive approach for screening or clinical diagnosis. In this work, we aim to leverage machine-learning principles to improve and validate the performance of a novel breath-based cancer screening tool.

Closing Skill Gaps in Under- and Unemployed Youth

Learning, an online learning platform, to underemployed or unemployed youth. Building on the work of a pilot project by CivicAction, the intern will plan and implement a randomized controlled trial to test factors driving uptake of online learning and its effects on employment outcomes. The intern and CivicAction will engage with Toronto-area employment services providers to provide various treatment intensities. Treatment groups range from a simple offering of a license to weekly reminders and curated playlists to encourage uptake or hands-on help in setting up online learning modules. Based on their pilot study, online learning has great potential to improve youth unemployment. This project will provide CivicAction with a rigorous evaluation of their online learning initiative. This evidence will inform future projects and support project expansion to larger populations.

Liability-driven investment strategy and portfolio immunization: a dynamicalapproach.

Since the recent financial meltdown and because of the current economic instability, many fund

managers tend to shift their total-return-oriented investment approach toward a liability-driven

investment (‘LOI’) strategy. The main objective of such a strategy is to find investment portfolios that

will match or outperform a liability stream (pensions, insurance claims, etc.). In a previous MITACS

project in collaboration with Addenda Capital, we investigated two basic liability matching strategies:

cashflow matching and moment matching. These strategies are now in use within funds managed by

Addenda Capital. The main objective of the present project is to further enhance the two methods.

The two LOI strategies are implemented by optimal selection of a portfolio of bonds (among a large

universe of liquid bonds) that will best match the cashflows or the moments of the liability stream. With

the passage of time the selected portfolio may no longer be optimal and may therefore require rebalancing.

We propose to implement multi-stage algorithms where the…..TOBECONTINUED

An Integrated model of Geomechanics and a Multiporosity Reservoir Simulator to Investigate Improved Recovery Techniques in Shale Reservoirs-Part 2

Shale reservoirs have become a very important source of hydrocarbons, especially in North America. Shales are rocks with very low permeability and therefore, produce the hydrocarbons stored in them is difficult. In order to do it, oil companies have to inject high pressurized fluids to break the rock. But, by using this unique strategy, most hydrocarbons are being left in the subsurface. This work aims to use mathematical and numerical models to investigate different methods that can lead to recover a bigger portion of the hydrocarbons stored in shale reservoirs. In this study, the application of the developed models will be focused on a particular shale reservoir, but the methodology can be applied to other fields in Canada

Need Definition and Need Mapping Methods for Innovation Projects

User needs analysis is a major concern in innovation projects and needs analysis methods differ in level of details, information overview, structure, interrelations, transparency, and ease of use. Depending on the project, the involved stakeholders, the required level of details, and time constraints, it is not clear which method should be used and why. The aim of this project is finding the best mapping method(s) and test them with the stakeholders. This project involves three phases. First, investigate current and future innovation projects in the organization to have an overview; prioritize and select three of them with different scopes. Second, investigate the work processes and human tasks; elicit the user needs and requirements to satisfy; select and use different methods to represent the needs. Third, test and compare different need mapping methods, or design a new one, by using them with different stakeholders to show their outcomes, strengths and limitations.