Federated Learning for Privacy-Enhanced Recommendations
Abstract
With the use of digital platforms increasing exponentially, the demand for personalized services emerges across a vast widespread of areas, from entertainment to finance. The adoption of these systems, regularly referred to as recommendation systems, may raise privacy concerns when implementations follow a centralized structure, approach that is commonly used. This dissertation focuses on exploring the validity of federated learning in the field of recommendation systems, specifically, in an e-commerce platform where the focus is privacy enhancement in addition to obtaining high recommendation accuracy. Within the scope of the Environment Adaptive Recommendation System project, an international researchdriven initiative focused on improving the current paradigm of recommendation systems, the present study will allow the integration of the developed methodologies, providing the project with a more strategically prepared foundation. After presenting the state of the art, the research firstly involves developing centralized models, followed by federated learning implementations to compare both approaches. In this stage, it is developed a solution that personalizes the recommendations of an e-commerce platform while the best practices are employed and different paradigms are considered. The study then evolves to the inclusion of privacy-enhancing mechanisms, to further study benefits and the situations their employment is relevant. The obtained results are then presented,…
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.