TY  - EJOU
AU  - Cai, Jincheng 
AU  - Feng, Li 
AU  - Zhao, Ni 

TI  - -FedVAE: Detached Posterior-Confidence Gating for Dimension-Wise KL Regularization in Personalized Federated Collaborative Filtering
T2  - Computers, Materials \& Continua

PY  - 2026
VL  - 89
IS  - 2
SN  - 1546-2226

AB  - Federated Variational Autoencoders (VAEs) keep interaction data local, but existing federated VAE recommenders typically apply uniform KL regularization and do not adapt dimension-wise penalties to unreliable posteriors in sparse interaction scenarios. We propose <math id="mml-ieqn-4"><mi>α</mi></math>-FedVAE, which uses a detached, clipped normalized signal-to-noise ratio as a local confidence gate for each KL dimension of a fused user posterior, without extra communication. Across MovieLens-100K, MovieLens-1M, and Amazon Video, <math id="mml-ieqn-5"><mi>α</mi></math>-FedVAE improves mean HR@20 by 7.8%–55.3% and NDCG@20 by 8.0%–65.8% over FedDAE. These results indicate that <math id="mml-ieqn-6"><mi>α</mi></math>-FedVAE improves personalized recommendation under sparse and decentralized settings while preserving the communication footprint.
KW  - Federated learning; self-adaptive training; variational autoencoder; collaborative filtering; regularization dynamics

DO  - 10.32604/cmc.2026.086183