Improving Federated Learning Accuracy with the Incremental Averaging Method: A Comparative Analysis of Model Aggregation Techniques

Given the increasing use of federated learning in the market and its promising future development prospects, this paper will mainly provide a new model merging method for federated learning called The Incremental Average method. After completing the local model construction using this method, customers can upload the model without waiting. This paper will first introduce the background, applications, and existing advantages and disadvantages of remote federated learning. It will then carefully analyze the current federated learning (FL) frameworks, including the formulas and methods involved. Following this, a new Incremental Averaging method will be presented and compared with the previous methods. The results show that the accuracy of remote federated learning significantly improves with the use of the Incremental Averaging method.