Distributed Online Learning and Stream Processing for a Smarter Planet

This chapter introduces the emerging paradigm of
stream processing and analysis
, including novel platforms and algorithms, that support the requirements of these kinds of applications. It also introduces distributed SPSs and proposes a novel distributed online learning framework that can be deployed on such systems to provide a solution to an illustrative Smarter Planet problem. The chapter defines a specific real‐world transportation inspired problem that requires large‐scale online learning. It then formalizes the characteristics of such problems and their associated challenges. The chapter discusses distributed systems and how the emergence of SPSs allows us to build and deploy appropriate solutions to these problems. It proposes a new framework for distributed, online ensemble learning that can naturally be deployed on a SPS to realize such applications, and describes how to apply such a framework to a collision detection application. The chapter concludes with a discussion on the several directions for future research enabled by this combination.