Causal Recommender Systems for Sequential Decision-Making

Recommender systems (RS) are intended to be a personalized decision support tool, where decisions can take the form of products to buy (e.g., Amazon), movies to watch (e.g., Netflix), online news to read (e.g., Google News), or even individuals to screen for a medical condition (e.g., personalized medicine). For digital users, RS play an essential role, since the available content (and hence possible actions) grows exponentially.

Lifelong reinforcement learning with autonomous inference of subtask dependencies

In this project, we propose a continual learning approach to face the problem of forward transfer in complex reinforcement learning tasks. Concretely, we propose a model that learns how to combine a series of general modules in a deep learning architecture, so that generalization emerges through the composition of those modules. This is of vital importance for Element AI to provide reusable solutions that scale with new data, without the need of learning a new model for every problem and improving the overall performance.

Improving Conditional Text Generation Algorithms

The aim of this project is to survey and develop methods for improving text generation algorithms. Natural Language Generation (NLG) is a subfield of Deep Learning that studies how to enable computers to write coherent texts. Current methods lack the capacity to adapt the content they generate, producing articles that are not always coherent nor factually correct. We will study ways of conditioning the algorithms to create more relevant content that adapts to the needs of the writer.

Apprentissage d’une pondération dynamique des constituants d’une méthode d’ensemble

L’utilisation de méthodes d’ensemble, dont l’usage est très répandu, permet souvent d’obtenir des résultats de haute précision mais sont beaucoup plus difficilement interprétable que les modèles d’apprentissage automatique traditionnel (arbres, régressions logistique/linéaire, ect).

Prédiction d’un indicateur de performance de ligne de production

Le passage des entreprises à l’industrie 4.0 à pour but de propulser la productivité, réduire considérablement
les coûts de production et d’améliorer grandement la qualité des produits.

Modeling Commodity Marketplace for Proof of Work Networks

Recent market instability, volatility, and Bitcion (BTC) halving event combined to create significant challenges for this sector, resulting in many hashing power producers being forced into bankruptcy. The sector has grown very rapidly and has been plagued with boom-bust cycles that have been difficult for producers to weather due to the lack of hedging tools/financial instruments at their disposal. Pow.re Corporation offers clearinghouse-type services providing hashing power producers the ability to sell their risk to speculators.

Optimisation des parcours de drones afin de maximiser la probabilité de détection d’une cible en détresse sur l’eau

Ce projet vise à optimiser les déplacements d’une flotte de drones qui sont utilisés pour détecter et reconnaître des personnes ou des bateaux sur une étendue d’eau permettant de minimiser les pertes humaines en intervenant le plus rapidement possible lors d’une situation d’urgence. La flotte de drones est reliée à une base et à une antenne 5G ce qui permet à l’équipe de secours de prendre des photos dans différentes zones de recherche.

Proposition de solutions en économie verte favorisant l’accessibilité des femmes à l’énergie

Le Centre d’Étude et de Coopération International (CECI) est une organisation favorisant l’innovation dans les projets de développement. Il souhaite aujourd’hui participer à la mise en œuvre de projets en économie verte dans les pays où il intervient afin de renforcer les capacités des communautés locales et notamment les femmes. Un de ses objectifs est de favoriser la coopération économique entre les pays d’intervention et le Canada.

Measuring firm sustainability using textual data

Finance has experienced a significant increase in studies tackling issues related to Sustainability in recent years. The majority of them rely on third-party ESG ratings which are diverse, not transparent, and lack standardization. Our project aims to use text-mining techniques to identify ESG dimensions in sustainability-related documents and offer an automatic measurement across the ESG dimensions. One objective of the generate text-based ESG scores will be then to establish correlations and trends between financial performances and Corporate Social Responsibility (CSR) programs.

Advanced Recommender Systems for Ecommerce

Numerous studies have recently proposed and highlighted novel techniques for recommendation, motivating the project of building the next generation of recommender systems for ecommerce platforms. This proposal aims at experimenting with one high-potential technique for modelling the recommendation problem, that makes great usage of the complexity of ecommerce data: Graph Neural Networks (GNNs). GNNs are increasingly used in recommender systems.

Pages