Applying FL to distributed learning for big data

In this section, we will discuss how FL can be applied to distributed learning in the context of big data.

FL for big data may not be related to privacy-related issues so much because the data needed for intelligence purposes is already possessed. Therefore, it may be more applicable to efficient learning for big data and improving training time significantly, as well as reducing the costs of using huge servers, computation, and storage.

There are several ways to conduct distributed learning on big data, such as building a specific end-to-end ML stack applied to different types of servers, such as parameter servers, or utilizing certain ML schemes on top of big data platforms such as Hadoop and Spark. There are also some other platforms, such as GraphLab and Pregel. You can use any libraries, and methods such as stochastic proximal descent and coordinate descent with low-level utilities for ML.

These frameworks can support the...