Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Alliance Grants

McGill University

$25,000.00

Lifelong Reinforcement Learning for Personalized Recommendation Systems

FY 2023-24

Summary
Recommendation systems are a core component of online e-commerce stores such as Amazon and eBay, online streaming services such as Netflix and YouTube, and social media platforms such Facebook and TikTok. Recommendation systems need to designed to balance multiple competing business objectives such as relevance of content to users, engagement and retention of users, revenue to the online platform, advertisement contracts with the sellers, etc. In many existing approaches, the recommendation systems are static: they are created once and used for several months or even years before a new recommendation system is designed, tested, and launched. Such static recommendation systems do not work well in times of market shock (such as the recent COVID pandemic) and changing user preferences. In this project, we propose to develop new theories, algorithms, and design guidelines to improve the robustness of recommendation systems to changing market conditions using ideas from lifelong reinforcement learning. Specifically, we propose to use recent advances in lifelong reinforcement learning and Dr. Mahajan's recent work on representation learning to develop a mathematically grounded framework for designing recommendation systems. We expect the resulting recommendation algorithm to continuously improve based on market conditions and changing user preferences. We will further exploit distributional reinforcement learning based recommendation system with robust transfer learning capabilities, so that the same recommendation algorithm can provide superior results in a wide variety of related businesses (such as same recommendation algorithm for music and movies). In addition, we will conduct extensive simulations to test the new recommendation algorithms for different use cases using public datasets such as Spotify music recommendation dataset, Retailrocket recommender system dataset, and Netflix Prize Data Set. It is anticipated that the project will lead to future publications in top-tier conferences and journals.
Program
Alliance Grants
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
580801-2022
Name nom
Mahajan, Aditya A
Department département
Electrical and Computer Engineering
Organization id
61
Institution établissement
McGill University
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2022
Award amount
25000
Program id
ALLRP
Program name en
Alliance Grants
Program name fr
Subventions Alliance
Group en
RESEARCH PARTNERSHIPS
Group fr
PARTENARIATS DE RECHERCHE
Committee code
96
Committee name en
RPP Internal Decision Cttee
Committee name fr
Cté de décision interne - PPR
Area of application code
1207
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Information, computer and communication technologies
Area of application fr
Technologie de l'information, des ordinateurs et des réseaux de communications
Research subject code
2805
Research subject group en
Artificial intelligence
Research subject group fr
Learning and inference theories
Research subject en
Intelligence artificielle
Research subject fr
Théories de l'apprentissage et l'inférence
Application title
Lifelong Reinforcement Learning for Personalized Recommendation Systems
Keyword
Artificial Intelligence; Lifelong reinforcement learning; Recommendation systems
Application summary
Recommendation systems are a core component of online e-commerce stores such as Amazon and eBay, online streaming services such as Netflix and YouTube, and social media platforms such Facebook and TikTok. Recommendation systems need to designed to balance multiple competing business objectives such as relevance of content to users, engagement and retention of users, revenue to the online platform, advertisement contracts with the sellers, etc. In many existing approaches, the recommendation systems are static: they are created once and used for several months or even years before a new recommendation system is designed, tested, and launched. Such static recommendation systems do not work well in times of market shock (such as the recent COVID pandemic) and changing user preferences. In this project, we propose to develop new theories, algorithms, and design guidelines to improve the robustness of recommendation systems to changing market conditions using ideas from lifelong reinforcement learning. Specifically, we propose to use recent advances in lifelong reinforcement learning and Dr. Mahajan's recent work on representation learning to develop a mathematically grounded framework for designing recommendation systems. We expect the resulting recommendation algorithm to continuously improve based on market conditions and changing user preferences. We will further exploit distributional reinforcement learning based recommendation system with robust transfer learning capabilities, so that the same recommendation algorithm can provide superior results in a wide variety of related businesses (such as same recommendation algorithm for music and movies). In addition, we will conduct extensive simulations to test the new recommendation algorithms for different use cases using public datasets such as Spotify music recommendation dataset, Retailrocket recommender system dataset, and Netflix Prize Data Set. It is anticipated that the project will lead to future publications in top-tier conferences and journals.