CLIP-based non-independent identically distributed data federal learning method

• Main Authors: WANG ZHIGUO, ZHU RUI, CHEN DONGSHENG, YIN KANGNING, JI XINHUI, WANG YAN, DING ZHEN
• Format: Patent
• Language: Chinese; English
• Published: 05.04.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention belongs to the technical field of federated learning, and discloses a non-independent identically distributed data federated learning method based on CLIP, which is suitable for model training between a server and a plurality of clients in communication connection with the server. Each client is provided with a CLIP model; the method comprises the following steps: firstly, a server determines a self-defined model and a training task, and sends the self-defined model and training parameters to all clients in communication connection with the server; each client uses the local training data set to train the custom model, and uploads the trained custom model to the server; and the server performs classification and aggregation according to the custom model from each client, generates a personalized custom model for each group, and distributes the personalized custom model to the corresponding client. According to the method, the user-defined model parameters uploaded by the client are grouped and t