Towards Accelerating the Adoption of Federated Learning for Heterogeneous Data

Federated Machine Learning (FML) is a distributed machine learning approach that solves basic AI and data problems such as data heterogeneity, privacy preservation, and data ownership. This technology enables organizations to collaborate on the model building while retaining control over their data, making it particularly useful when data is sensitive or too large to be collected in a central location. Numerous open-source frameworks for FML have been developed, each with different capabilities. In this paper, we use a popular framework to implement a proposed algorithm and tackle the significant problem of data heterogeneity in AI. Specifically, we integrated the FEDMA algorithm to simulate the data heterogeneity problem with the FEMNIST dataset, a widely used benchmark dataset in the research community.