A Survey Report on TinyML and Small Data – the future of machine learning

This survey report provides a comprehensive analysis of the concepts of Small Data and Tiny Machine Learning (TinyML), along with their respective challenges, techniques, and applications. Small data refers to datasets that are relatively limited in size, typically characterized by a low number of samples, sparse feature space, or data scarcity due to privacy concerns or data collection challenges, while TinyML is a field of study in Machine Learning and Embedded Systems that explores the types of models you can run on small, low-powered devices like microcontrollers. It enables low-latency, low power and low bandwidth model inference at edge devices. This report aims to explore the landscape of small data and TinyML, shedding light on the unique challenges they present and the strategies employed to tackle them. The report begins by defining small data and highlighting its significance in domains where large-scale data collection is impractical or unfeasible. It explores the characteristics of small data, such as sparsity, noise, and class imbalance. Subsequently, the report delves into the emerging field of TinyML, which seeks to deploy machine learning models on devices with limited resources, such as IoT devices. It examines the challenges inherent to TinyML, including limitations computational power and memory, and energy constraints, as well as the unique considerations involved in training and deploying models on such devices. Furthermore, the report explores the diverse applications of small data and TinyML across various domains, such as healthcare, marketing, retail, agriculture, and manufacturing. It showcases real-world use cases that leverage small data and TinyML techniques to solve critical challenges and enable intelligent decision-making at the edge. By presenting a comprehensive survey of small data and TinyML, this report provides researchers, practitioners, and decision-makers with valuable insights into the potential and limitations of these fields. It offers a foundation for further exploration, innovation, and advancements in small data analysis techniques and the deployment of machine learning on resource-constrained devices.