Software Framework for Tiny Machine Learning
Abstract
This chapter reviews the evolutionary path and future perspectives of software frameworks for tiny machine learning (TinyML), which enable the computation of innovative machine learning (ML) models on smaller, commodity devices. The chapter provides an introductory, concise overview of TinyML, its role in the Internet of Things (IoT), the edge computing era, and its global significance. It then addresses the unique challenges of building and deploying ML models on hardware with limited processing power, memory, and energy. This chapter provides an overview of commonly used software frameworks that address these challenges, outlining their application domains, architectural elements, and design principles. The chapter also emphasizes the solutions these frameworks provide for model development, training, and deployment. The current trends, research, and future directions within TinyML software frameworks, which hold promise for transforming edge AI applications in healthcare, agriculture, and industrial IoT, are also presented.
