Kategoria BISAC: Computers >> Embedded Computer Systems

This book describes machine learning-based new principles, methods of design and optimization of high-speed integrated circuits, included in one electronic system, which can exchange information between each other up to 128/256/512 Gbps speed. The efficiency of methods has been proven and is described on the examples of practical designs. This will enable readers to use them in similar electronic system designs. The author demonstrates newly developed principles and methods to accelerate communication between ICs, working in non-standard operating conditions, considering signal deviation...

This book provides readers a comprehensive introduction to alternative number systems for more efficient representations of Deep Neural Network (DNN) data. Various number systems (conventional/unconventional) exploited for DNNs are discussed, including Floating Point (FP), Fixed Point (FXP), Logarithmic Number System (LNS), Residue Number System (RNS), Block Floating Point Number System (BFP), Dynamic Fixed-Point Number System (DFXP) and Posit Number System (PNS). The authors explore the impact of these number systems on the performance and hardware design of DNNs, highlighting...

This book presents recent advances towards the goal of enabling efficient implementation of machine learning models on resource-constrained systems, covering different application domains. The focus is on presenting interesting and new use cases of applying machine learning to innovative application domains, exploring the efficient hardware design of efficient machine learning accelerators, memory optimization techniques, illustrating model compression and neural architecture search techniques for energy-efficient and fast execution on resource-constrained hardware platforms,...