Название: Minimizing Data Movement and Parameter Count Across the Machine Learning Stack: Everything is a Matrix
Автор: Andrew Sabot
Издательство: Springer
Серия: Synthesis Lectures on Computer Science
Год: 2026
Страниц: 121
Язык: английский
Формат: pdf (true), epub
Размер: 16.6 MB

This book provides a focused, research-forward guide to making large AI models efficient in practice and also presents an array of novel techniques to reduce memory footprint, accelerate computation, and improve overall hardware utilization. The author demonstrates that substantial efficiency gains can be achieved by rethinking how data is computed, stored, and compressed, with a special focus on matrices, the core computational structure underpinning both scientific computing and neural networks. Modern AI models run on huge grids of numbers (matrices/tensors), and their speed and affordability depend on how those numbers are arranged and processed on real hardware (GPUs/TPUs/CPUs). This book explains practical methods to skip unnecessary work (structured sparsity), move data efficiently (gather/scatter), and shrink models without losing accuracy (block distillation) so that AI systems can use less memory, less time, and less energy without sacrificing quality. In addition, the book shows how to turn algorithmic ideas into hardware-aware speedups on GPUs/TPUs. Readers will learn when sparsity pays off, how to schedule irregular workloads, and how to recover accuracy in compressed models. Case studies illustrate end-to-end design choices, evaluation, and pitfalls. The result is a coherent perspective that bridges theory, compilers/run times, and real-world deployment.

The rapid ascent of Machine Learning has redefined the boundaries of what is computationally possible, from real-time natural language understanding to sophisticated computer vision. However, as the field of Artificial Intelligence continues to progress, the industry has reached a critical juncture: the sheer scale of modern models has outpaced the growth of the hardware required to run them. The “computational tax” of Artificial Intelligence is no longer just a technical hurdle; it is a sustainability and accessibility crisis.

This book is born out of the necessity to bridge the widening gap between algorithmic ambition and hardware reality. At its core, it argues that the next leap in AI will not come solely from larger datasets or deeper layers but from fundamental innovations in how we manage the matrix operations that underpin a substantial amount of modern workloads.

Target Audience:
This book is intended for researchers, high-performance computing (HPC) engineers, and graduate students who are looking for more than just a theoretical understanding of AI. It provides actionable techniques for those tasked with deploying large-scale models in resource-constrained environments. This book includes optimization strategies ranging from on-device edge computing to massive, energy-conscious data centers.

Contents:

Ссылки удалены по требованию правообладателя

ОТСУТСТВУЕТ ССЫЛКА/ НЕ РАБОЧАЯ ССЫЛКА ЕСТЬ РЕШЕНИЕ, ПИШЕМ СЮДА!