Scenarios for Federated AI
Dinesh Chandra Verma · 2021
Most real-world situations in businesses deal with systems that are distributed over a wide area and interconnected by a computer communications network. The different components may be located in different parts of the world connected by the Internet, or be within different sites of a company connected by the corporate intranet. These sites can be defined into three categories - edge sites, proxy sites and central site. Each of these sites could have different roles including generating data, collecting data, training AI models from data, drawing inference from data using an AI model, or acting upon the inference. The assignments of the roles to edge, proxy and central site results in different patterns for enterprise AI. This chapter reviews the patterns of enterprise AI, outlining the role of each category of the site in each stage of the AI life-cycle model. It discusses the motivation for patterns that require federation at different stages of the AI life-cycle, and outlines the differences between consumer federated learning and enterprise federated learning. For enterprise federated learning, several scenarios where federated learning would be required are described.