Название: Artificial Intelligence and Hardware Accelerators
Автор: Ashutosh Mishra, Jaekwang Cha, Hyunbin Park
Издательство: Springer
Год: 2023
Страниц: 358
Язык: английский
Формат: pdf (true)
Размер: 17.1 MB

This book explores new methods, architectures, tools, and algorithms for Artificial Intelligence Hardware Accelerators. The authors have structured the material to simplify readers’ journey toward understanding the aspects of designing hardware accelerators, complex AI algorithms, and their computational requirements, along with the multifaceted applications. Coverage focuses broadly on the hardware aspects of training, inference, mobile devices, and autonomous vehicles (AVs) based AI accelerators.

Smart technologies and an intelligent society are the demand of this era. Artificial Intelligence (AI) and Deep Learning (DL) algorithms are playing a vital role to cater these demands to meet the expectations of the smart world and intelligent systems. Increased computing power, sensor data, and improved AI algorithms are driving the trend toward Machine Learning (ML) in cloud-based and edge-based intelligence. Both are applicable through intelligent devices, wearable electronics, smartphones, automobiles, robots, drones, etc. However, efficient hardware can solely accomplish the required performance of implementing these algorithms. Therefore, AI accelerators are the state-of-the-art research area for circuits and systems designers and academicians. AI accelerators are needed to cater to the insatiable demands of compute-intensive AI applications.

Artificial Intelligence (AI) is designing new genesis around the globe and garnering great attention from industries and academia. AI algorithms are indigenously intensely computational and data ambitious, which entails appropriate AI accelerators to comply with the computational efficiency. In general, AI professionals and hardware designers are different persons having no expertise in either area. This gap between the algorithmic and hardware experts brings an intense demand for a bridge to subdue this technological slit. This book aims to bridge this gap by assimilating state-of-the-art technologies and architecture. The chapters are arranged by considering computing system platforms, technological development, emerging applications, and the existing issues in AI hardware accelerators.

Chapter 1 comprehends the basics of AI, algorithms, and processor designs. It specifies the demands and requirements of AI and the hardware. It summarizes the entire book in a glance and induces the desire for an in-depth understanding of the technologies involved in recent accelerator designs. The computing system platforms are considered in Chaps. 2, 3, 4, 5, and 6. Chapter 2 deals with “single-tenant,” i.e., single owner of the compute. The standalone personal computer is an example of single-tenant computing. This chapter describes CPU and GPU-based accelerators, including training and inference accelerators. Chapter 3 incorporates “multiple tenants,” i.e., the owners associated with the servers or cloud computing independently. It also includes training and inference accelerations. Chapters 4 and 5 are dedicated to the AI accelerators in smartphones. Further, Chap. 6 provides the AI acceleration for embedded computing systems.

The trending technological developments in the AI accelerators have been considered in Chaps. 7 and 8. Chapters 7 consists of the reconfigurable and programmable AI accelerators (FPGA and ASIC dedicated hardware accelerators). The neuromorphic approach of the hardware design for AI accelerations is considered in Chap. 8. It discusses the state-of-the-art devices, circuits, and architectures used in neuromorphic computing. An emerging application of AI accelerators is elicited in Chap. 9. It contains accelerators for AVs. It describes AV computing platforms such as NVIDIA Jetson and Tesla’s full self-driving (FSD) computer. Energy is a crucial parameter for designing any efficient hardware. The AI models are highly hungry for energy due to extensive calculations and voluminous data. Chapter 10 includes the demand for energy in an AI accelerator.

Every chapter presents the requirements of hardware corresponding to their theme and provides comprehensive explanations of the recently available solutions and developments. In this way, the book will be helpful to design engineers, researchers, hardware experts, and industry employees to obtain insight into the AI models and algorithms. Therefore, they can develop an enhanced and efficient computing platform to accelerate AI applications and implement intelligence in devices. This volume can also serve as a textbook or reference book for undergraduate and graduate students in either domain (AI or hardware). The literature and survey covered in this book are very recent and relevant to the theme of the corresponding chapter. Therefore, a refined and inclusive survey of the AI hardware accelerator designs is encapsulated in this book, and it will be an add-on advantage to its readers.

Contents: