Optimization of Federated Learning Communication Costs through the Implementation of Cheetah Optimization Algorithm

Recently, Federated Learning (FL) has promptly gained aggregate interest owing to its emphasis on the data privacy of the user. As a privacy-preserving distributed learning algorithm, FL enables multiple parties to construct machine learning (ML) algorithms without exposing sensitive information. The distributed computation of FL may lead to drawn-out learning and constrained communication processes, which necessitate client-server communication cost optimization. The two hyperparameters that have a considerable effect on the FL performance are the number of local training passes and the ratio of chosen clients. Owing to training preference across different applications, it is challenging for the FL practitioner to manually choose these hyperparameters. Even though FL has resolved the problem of collaboration without compromising privacy, it has a transmission overhead because of repetitive model updating during training. Various researchers have introduced transmission-effective FL techniques for addressing these issues, but sufficient solutions are still lacking in cases where parties are in charge of data features. Therefore, this study develops an Optimization of Federated Learning Communication Costs through the Implementation of the Cheetah Optimization Algorithm (OFLCC-COA) technique. The OFLCC-COA technique is mainly applied for effectually optimizing the communication process in the FL to minimize the data transmission cost with the guarantee of enhanced model accuracy. The OFLCC-COA technique enhances the robust performance in unsteady network environment via the transmission of score values instead of large weights. Besides, the OFLCC-COA technique improves the communication efficiency of the network by transforming the form of data that clients send to servers. The performance analysis of the OFLCC-COA model occurs utilizing different performance measures. The simulation outcomes indicated that the OFLCC-COA model obtains superior performances over other methods in terms of distinct metrics