Summary

This chapter provided an overview of how FL could potentially solve many of the big data issues by first understanding the definition of big data and its nature, involving an abundance of observations, acceptance of messiness, and ambivalence of causality.

We have learned about privacy regulations in a variety of forms from many regions and the risk of data breaches and privacy violations that eventually lead to loss of profits, as well as a bottleneck in creating authentic AI applications. Federated learning, by design, will not collect any raw data and can preserve data privacy and follow those regulations.

In addition, with an FL framework, we can reduce inherent bias that affects the performance of ML models and minimize model drift with a continuous learning framework. Thus, a distributed and collaborative learning framework such as FL is required for a more cost-effective and efficient approach based on FL.

This introductory chapter concluded with the potential of FL as a primary solution for the aforementioned big data problems based on the paradigm-shifting idea of collective intelligence that could potentially replace the current mainstream data-centric platforms.

In the next chapter, we will see where in the landscape of data science FL fits and how it can open a new era of ML.