Federated Hash Projected Latent Factor Learning

Short summary

New federated learning method combines hash learning with privacy-preserving distributed training, replacing real-valued gradients with efficient binary representations to reduce communication overhead. Introduces Projected Hamming Distance for improved similarity modeling and includes a secure gradient reassembly strategy. Outperforms state-of-the-art methods across four real-world datasets.

• •Combines hash learning with federated learning for privacy-preserving binary representations
• •Uses binary gradients instead of real-valued ones, significantly cutting communication and computation costs
• •Outperforms existing approaches on four datasets while improving privacy, efficiency, and model accuracy

Generated with AI, which can make mistakes.