From Human Attention to Computational Attention: A Multidisciplinary Approach by Matei Mancas, Vincent P. Ferrera, Antoine Coutrot ISBN 9783031842993, 9783031843006, 3031842995, 3031843002 instant download
The purpose of this book is to present a multi-disciplinary perspective on the modelling of human attention which is of great interest for artificial intelligence (AI).
This second edition of the book delves into the arrival of deep learning, which has influenced attention models and has in turn been influenced by attention, particularly in architectures such as transformers. It also presents more work on the neuroscience side and emphasises how neuroscience can inform the AI domain and vice versa.
The book structure is organised around four parts which are detailed in Chap. 1. The first part, called “Foundations,” is organised around three chapters, focuses on fundamentals and is a comprehensive introduction to attention modelling.
The second part, called “Attention in the Brain,” is organised around three chapters. It deals with neuroscience and details where attention takes place in the brain, how neurophysiology can inform signal detection and how to fill the gap between the study of a single neuron and higher level tasks such as visual performance.
The third part, called “Attention in Computer Science,” is organised around six chapters and focuses first on how attention is used in engineering, model validation, how attention can be applied on multimodal data and finally a chapter on attention in deep learning architectures and especially in transformers.The fourth part, called “Convergence: when the brain informs computer science (end vice versa),“ organised around three chapters contains first a chapter on how to inform brain research from results in computer science and describes a theory of information seeking in the brain which can explain practical implications of attention. Perspectives on the different fields of attention conclude the book.
This book intends to provide important information for both students and researchers of attention with approaches ranging from engineering to neuroscience domains.
*Free conversion of into popular formats such as PDF, DOCX, DOC, AZW, EPUB, and MOBI after payment.