Distributed learning nature – toward scalable AI

In this section, we introduce the distributed computing setting and discuss the intersection of this setting with ML approaches to fully establish the support for why FL is necessary. The goal of the section is for the user to understand both the benefits and limitations imposed by the distributed computing setting, in order to understand how FL addresses some of these limitations.

Distributed computing

The past several years have shown a large but predictable rise in the development of new approaches and the conversion of existing server infrastructure within the lens of distributed computing. To generalize further, distributed approaches themselves have shifted more and more from research implementations to extensive use in production settings; one significant example of this phenomenon is the usage of cloud computing platforms such as AWS from Amazon, Google Cloud Platform (GCP) from Google, and Azure from Microsoft....