A comprehensive survey of federated learning and recommender system

What is it about?

Faced with the urgent challenge of guaranteeing recommendation performance and personal data security simultaneously, Federated Learning (FL) enters researchers' vision. The critical point of FL is that users' private data are retained locally and not allowed to be shared. It provides a new perspective for privacy-preserving recommendation and facilitates cross-field research named "Federated Recommender System (FedRec)." The goal of FedRec is to provide users with reasonable recommendations without privacy violations. We propose a hierarchical taxonomy to summarize related work from model, privacy and federated perspectives systematically. The taxonomy not only highlights current innovation in federated recommendation algorithms, but also emphasizes the importance of privacy protection.

Featured Image

Photo by Tushar Mahajan on Unsplash

Photo by Tushar Mahajan on Unsplash

Why is it important?

(1) From the model perspective, we group federated recommendation papers into different learning paradigms (e.g., deep learning and meta learning). (2) From the privacy perspective, privacy-preserving techniques are systematically organized (e.g., homomorphic encryption and differential privacy). (3) From the federated perspective, fundamental issues (e.g., communication and fairness) are discussed. (4) Each perspective has detailed subcategories, and we specifically state their unique challenges with the observation of current progress. (5) Finally, we figure out potential issues and promising directions for future research.

Perspectives