Kuzdeba S. Radio Frequency Machine Learning: A Practical Deep Learning Perspective

Artech House Publishers, 2025. — 259 p. — ISBN 1685690335.This book is a one-stop guide for wireless enthusiasts, serious researchers, interested students, or course instructors who are interested in the question of “what next ?” It aims to demystify the evolution of wireless along two important axes, with (1) massive data and computation ability and (2) traditional wireless optimization metrics. As an example, graphics processing unit (GPU) clusters can exploit parallelism in computing, leading to teraflops of processing power in the form factor of a smartphone. Several terabytes of data can be captured in a few minutes with a software defined radio that fits in the palm of your hand. When incredible computation at the network edge meets abundant data collection capability, there is an unprecedented opportunity to utilize one of the most promising tools that has taken the last decade by storm—machine learning—in context of both deepening our understanding of wireless as well as optimizing and tuning wireless networks without immediate human intervention. While classical and domain-focused texts on the basic principles of wireless communications and networking will always be needed to lay down the foundations of the discipline, there is a clear and well-defined need to harness all the resources in machine learning that an engineer can access today, which was unthinkable even a decade ago. This book is hence timely and is designed to arm the reader with not only theory and concepts but also use cases and examples to illustrate when machine learning can be effective, and alternatively, when traditional signal processing should be preferred. At its core, it distills the hype from the promise, which is truly a need within the community to advance adoption of machine learning tools in practical wireless deployments.
This book first introduces the foundational concepts in Chapter 1, highlighting the shift from traditional to machine learning-based processing at the physical layer. In Chapter 2, it dives into machine learning-aided classification of emitters and signals, explaining different models and architectures with their salient features that make them uniquely qualified for wireless use cases. Chapter 3 covers clustering and visualization, looking at how to address learning when you do not have labeled datasets. Chapter 4 tackles the exciting new area of generative machine learning, which allows data augmentation and waveform synthesis. After thoroughly establishing the algorithmic and theoretical principles thus far, Chapter 5 focuses on designing systems that can accept streaming data for real-time operations as well as iterative methods of reinforcement learning and fusing of multimodal data. To overcome the barrier of entry and deployment in practical systems, Chapter 6 is dedicated towards robust training of such models and touches on the key issue of trust and explainability. The success of machine learning models depends on the quality and accessibility of data, and new advances in data augmentation and synthesis are covered in Chapter 7. Finally, Chapter 8 looks forward and identifies other essential emerging concepts that will undoubtedly unlock many new capabilities in this design and implementation space.True PDF