p01 / Journal
Federated Neural Collaborative Filtering
Knowledge-Based Systems, 242, 108441 · 2022
In plain language
Federated recommendations without collecting users’ interaction histories in one place. FedNCF adapts neural collaborative filtering to collaborative local training.
Research context
Recommendation quality can be maintained while reducing the need to collect raw personal data.
Approach
Neural collaborative filtering, privacy-preserving aggregation, and separate averaging rules for latent factors and neural weights.
Findings
The experiments report recommendation quality comparable to centralized NCF and faster convergence with the proposed aggregation rule.
Summary based on abstract.