Metaheuristic Approaches to Feature Engineering and Model Architecture Optimization for Financial Time Series Prediction

Predictive modeling of financial data, especially trading activity or asset prices, is a very challenging task. There are a number of novel approaches to feature engineering, data preparation and model architectures that aim to mitigate some of the problems that arise from non-stationarity and other issues typically found in financial time series data.

Identifiability of latent factors through multiple self-supervision

Human perception has developed the ability to decompose scenes into fine grained elements. This lays the foundation for strong generalization to new situations where the base concepts can be recomposed to interpret objects never seen before. While it has been shown that, in the general case, proper decomposition is not possible, new paradigms provide provable decomposition in constrained environments. We hypothesize that the multiple sensory systems of human perception offer a strong signal for decomposing scenes in a proper way.

Optimization of sentence classification for insurance applications

This research is carried out on the topic of natural language processing and specifically on word representation on Question Answering tasks. The state of the art in Question Answer task is Google’s Bidirectional Encoder Representations from Transformers (BERT) language model.
Koïos Intelligence is interested in fine-tuning this model for their closed domain artificial intelligence (AI) virtual assistant, targeted at insurance and financial applications.

Discount Pricing Recommendations

It is well known that retailers have razor-thin margins. A few discount percentage points can make the difference between a bad and an excellent year. The goal of this project is to make sure that Altitude Sports’ prices are optimized to satisfy customers and maximize margins, all year long. There are a lot of factors influencing pricing decisions such as official and unofficial MAP (Minimum Price Policies enforced by brands), stock velocity, season stock levels, stock scarcity, cashflow and price elasticity. Furthermore, price elasticity in itself is highly variable.

Computer Vision for Crop Weed Identification

Aid in the development of a Machine Learning Model for utilization by an Agricultural Robot. This Robot performs several tasks, primarily the mechanical removal of weeds from vegetable farms. Therefore, the machine learning model is concerned with informing the robot of locations of interest points of the weed and crop plants, as viewed from several sensors mounted on the robot. Other sensors of the robot, such as GPS, and wheel odometry, can be brought to bear as well.

Natural Language Processing for Automated Classification and Analysis of Aviation Safety Reports

The aviation industry connects people, markets, and cultures around the world and aviation is the key to ensuring that air transport continues to play a major role in driving sustainable economic and social development.

Briser l’isolement des ainés en temps de COVID-19 grâce au système d’échange local mis en place par l’Accorderie de Sherbrooke

L’urgence sanitaire liée à la COVID-19 a forcé les ainés à adopter des mesures drastiques de confinement qui les mettent à risque d’un isolement. À long terme, ce confinement peut conduire à un désintérêt de la vie. L’organisme communautaire l’Accorderie de Sherbrooke, qui propose un système d’échange local entre ses membres, a mis sur pied dès le début du confinement un programme de jumelage entre un membre et un ainé pour les livraisons à domicile, générant 120 nouveaux membres.

Discovery of Root Causes of Quality Deviations in Electronics Manufacturing

A typical production line in electronics generates an important quantity of data which is generally ignored and unused. Tack Verification increases efficiency of electronics manufacturing companies by collecting and transforming operational data into actionable insights and key performance indicators.
The project consists of the research and primary validation of artificial intelligence models allowing discovery of the root causes of quality deviations measured during electronics manufacturing.

COVID-19 : module pharmaceutique pour le développement accéléré de produits désinfectants

Galenit est une plateforme de formulation qui facilite le développement de produits de santé. Dans le cadre de ce projet, Galenit souhaite développer sa plateforme afin d’y intégrer l’aide à la formulation de produits désinfectants.

Decentralized Deep Radiomics: Scaling up the discovery of prognostic and predictive cancer imaging biomarkers from routine clinical data across a network of hospitals

Genetic advances over the past 10 years have led to the development of several targeted therapies for lung, breast and colon cancer. However, there are a number of factors that limit the optimal use of these innovations, including the high cost of the organizational process associated with molecular testing, and their late use in the patient's journey. Recently, the prospect of obtaining non-invasive, cost-effective and timely triggers for diagnostic & therapy has emerged from a discipline known as Radiomics.