一种联邦迁移学习方法、装置、存储介质及电子设备
By identifying common sample objects in federated transfer learning and using a third-party server to calculate feature loss to train and extract subnets, the complexity and time-consuming nature of existing technologies are solved, achieving a more efficient training process.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ALIPAY (HANGZHOU) INFORMATION TECH CO LTD
- Filing Date
- 2023-05-26
- Publication Date
- 2026-07-17
AI Technical Summary
The existing federated transfer learning process is complex and lengthy, consuming a lot of time and computing resources, especially when considering privacy protection, as both parties need to exchange data through multiple rounds of communication.
By identifying the common sample objects of the participants as the target sample objects, their data is input into the extraction subnet to be trained. A third-party server is used to calculate the feature loss between the feature to be optimized and the standard feature, and the extraction subnet is trained based on this, reducing the number of communication rounds and the amount of computation.
Training the extracted subnets is completed in fewer communication rounds, which significantly reduces the cost and computational load of federated transfer learning and improves efficiency.
Smart Images
Figure CN116629381B_ABST