Certain investigations on distributed learning performance enhancements in federated multitask learning

Abstract

Federated learning enables machine learning models to learn from newlinedecentralized data without compromising privacy. The standard formulation newlineof federated learning produces a shared model for all users. Due to statistical newlineheterogeneity and the non-IID distribution of data across devices often by newlineusers, the local models trained solely on their private data perform better than newlinethe global shared model, which will take away their incentive to participate in newlinethe process. Several techniques have been proposed to personalize global newlinemodels to work better for individual users. As solutions, personalized newlinefederated learning and Federated Multitask Learning (FMTL) have been newlineproposed to handle the statistical diversity in FL. Personalized FL aims to newlinebuild a global model, which should be an advantage to finding a personalized newlinemodel that is stately for each users data. Here, the global model considered an newlineagreed point for each user to start personalizing its local model based on its newlineheterogeneous data distribution. FMTL directly addresses the challenge of newlinestatistical diversity in FL by learning simultaneously separate models for each newlineuser. In FMTL, the users models are separated but typically correlated, since newlineusers with similar features are likely to share similar behaviors. The newlinerelationships among the users models are to be captured by a regularization newlineterm, which is minimized to encourage the correlated users models to be newlinemutually impacted. Constituting a large amount of disease-related data from newlineheterogeneous devices in personalized models can be learned by using newlineFederated Multitask Learning (FMTL). Due to system and statistical newlineheterogeneity, a personalized model has been studied by Federated Multitask newlineLearning (FMTL) to predict the updated infection rate of COVID-19 in the newlineUSA using a mobility-based SEIR model. Mobility-based SEIR model with newlinean additional constraint, we can analyze the availability of beds. newline