FedKLPR: KL-Guided Pruning-Aware Federated Learning for Person Re-Identification

In three linesFedKLPR is a federated learning framework for person re-identification combining KL-Divergence-guided training to handle statistical heterogeneity, unstructured pruning for communication efficiency (40-42% reduction on ResNet-50), and Cross-Round Recovery for adaptive compression control. Evaluated on 8 benchmark datasets.

Summary generated by Claude — human-verified