Supporting post-WIMP interactions in Mixed Reality with Spatial Computing and Edge Computing

Mixed Reality (MR) allows the display of virtual content, for instance with holograms, that can merge to variable degrees with the physical environment of a user by taking into account the context and topology of the user’s environment. By its mobile, spatial and context-aware nature, MR allows to interact with virtual content with new interaction paradigms that are closer to the natural interactions we have with everyday objects.

Transition model for insurance risks

Car (automobile) insurance is a very common type of insurance: policyholders pay a premium to get financial compensation in case an accident happens with their cars. Insurance companies use complex calculations and a lot of information to determine the value of these premiums. More specifically, they must also consider their expectations of the future. Predicting the future is impossible but with the help of artificial intelligence, the current project aims to improve the understanding of how a portfolio of insured cars can evolve in the upcoming years.

ESG Events Clustering using Natural Language Processing

The integration of Environmental, Social and Governance (ESG) factors into investment decisions has accelerated in the last few years. In fact, Bank of America estimates a $20 trillion of asset growth in ESG funds over the next two decades. Evaluating ESG related events of a company is an important task to assess company ESG risk. For public companies, important changes should be covered by the media and possibly by several news. These news need to be grouped under the same ESG event before analysis.

Composing without forgetting

In this project, we propose a modular continual learning approach to face the problem of catastrophic forgettingand transfer in learning from evolving task distributions. Concretely, we propose a model that learns how to selectmost relevant modules based on a local decision rule for a given task to form a deep learning model for solving agiven task. In this framework we generalization to unseen but related tasks emerge through the composition ofthose modules.

Assistance tools to build taxonomy of skills and to search for experts

In many applications, we need to find the right person with the required expertise (skills) for some task. To enable this, it is necessary to extract and organize the skills of different persons in a taxonomy. This project aims at developing a taxonomy of skills for the National Bank of Canada. Currently, a small taxonomy has been created manually by experts, but it is far from being complete. Many skills and their variants are missing. The project will construct tools to extract skills from the profiles of the employees.

Développement d'une méthode non-supervisée d'apprentissage profond pour la détection de défaillance à partir de signaux acoustiques et vibratoires

L'objectif du projet est d'automatiser la reconnaissance des signaux acoustiques et vibratoires a l'aide d'un modele d'apprentissage profond. Cette reconnaissance est la premiere etape du developpement d'un modele de pronostic et de suivi de !'evolution de la degradation des alternateurs. Ces signaux sont issus de mesures effectuees periodiquement sur les alternateurs durant leur fonctionnement.

Apprentissage par renforcement pour la planification de la production dans le réseau électrique québécois de 2035

À Hydro-Québec, le « Plan de production » est la stratégie de gestion du parc de centrales mise en place afin de répondre à la demande en électricité à court terme (horizon de 30 jours) et à très court terme (horizon de 24 heures). L’objectif du projet est d’optimiser la prise de décision sur les horizons très court terme et court terme sur le réseau de transport québécois de 2035.

Artificial intelligence for automated identification of cannabaceae family plant diseases and gender on low-resource devices

Automated and early identification of plant diseases from their leaves is an important task in agriculture and can have positive impacts on crop yield and quality. Crop pathogens and pests reduce the yield and quality of agricultural production. They cause substantial economic losses and have a direct impact on food security and nutrition worldwide. Due to the wide variety of crops and diseases, even a farmer or pathologist can often fail to identify plant diseases by visualizing the affected leaves. However, visual observation remains the primary approach to disease identification.

Automating the development of new cosmetic formulations through machine learning.

The cosmetics industry represents a market with sales of more than US 1 0 billion worldwide. It is estimated that sales of cosmetic products are expected to grow annually by more than 3.7% until 2022. Extremely dynamic with a development cycle of 1 months, this sector is also extremely regulated at the national level without any international harmonization. These aspects constitute a significant obstacle to the development of new products for VSEs and SMEs in the sector, thus greatly reducing their development capacity.

Preuve de concept d’un système d’aide à la navigation pour le transport maritime afin de réduire l’impact du bruit dans l’environnement du Bélugas

Les technologies liées aux systèmes d’aide aux prises des décisions dans des navires intelligents ont atteint un niveaude maturité élevé dans les dernières années. Pendant ce temps, les navires autonomes et sans pilote ont aussi étélargement étudiés en parallèle aux véhicules autonomes, comme les chargeuses autonomes dans les mines.