Federated Learning Acceleration Patent for Intelligent Personalized Services

Korea Electronics Technology Institute filed patent application US20260094005A1 for a method of accelerating federated learning client training. The invention covers searching models with similarity above a predetermined threshold in a model repository based on user requests and embedded data, generating a weighted global model from searched models, distributing the global model to user and participating clients for training, and training the global model to reflect pre-stored local data. The patent is classified under CPC G06N 3/098.

This patent application does not create immediate compliance obligations for third parties. Technology companies developing federated learning systems, AI personalization engines, or distributed machine learning training infrastructure should review the published claims to assess potential patent landscape implications. The application represents a draft stage intellectual property claim that has not yet been examined or granted.