Fig. 6
From: Federated cross-view e-commerce recommendation based on feature rescaling
Performance comparison of recommendation models built with the Fed-FR-MVD framework alongside centralized and federated methods. This figure aims to illustrate the varying convergence trends and final performance metrics of different models over training iterations, highlighting the strengths of the Fed-FR-MVD approach in achieving superior stability and effectiveness in Recall and NDCG, especially in later training stages, compared to the faster initial convergence of other federated methods like Fed-SV. The results underscore the potential advantages of adopting the Fed-FR-MVD framework for improved recommendation performance in federated learning contexts.
