Self-attention mechanism and noise reduction techniques combined with federated comparative learning feature aggregation algorithm
The application of traditional artificial intelligence(Al)techniques in the field of intelligent vision faces a series of technical difficulties such as data heterogeneous distribution,data privacy protection,and noisy data interference.To o-vercome these challenges,this paper proposes a comparative federated learning aggregation feature algorithm with noise re-duction(FedCAFNR),based on the ideas of federated learning and signal noise reduction algorithms.This paper adopts a self-attentive mechanism and a deep residual network to build an adaptive threshold noise reduction network,thus improving the ability of federated learning to learn key features from heterogeneous data.This paper explores the differences between comparing different learning models and optimizes local training to further improve the consistency of the models.The pro-posed algorithm is evaluated on three multi-classification datasets and simulation experiments demonstrate that FedCAFNR has good convergence and scalability while ensuring data privacy.In addition,the new algorithm improves the learning effi-ciency and increases the consistency and robustness of existing algorithms for federal learning in heterogeneous data environ-ments.
万方数据
维普
