Название: Interpretable Artificial Intelligence: A Perspective of Granular ComputingАвтор: Witold Pedrycz, Shyi-Ming Chen
Издательство: Springer
Год: 2021
Страниц: 430
Язык: английский
Формат: pdf (true)
Размер: 15.1 MB
In the recent years, Artificial Intelligence (AI) has emerged as an important, timely, and far reaching research discipline with a plethora of advanced methodologies and innovative applications. With the rapid progress of AI concepts and methods, there is also a recent trend to augment the paradigm by bringing aspects of interpretability and explainability. With the ever growing complexity of AI constructs, their relationships with data analytics (and inherent danger of cyberattacks and adversarial data) and the omnipresence of demanding applications in various critical domains, there is a growing need to associate the results with sound explanations and augment them with a “what-if” analysis and advanced visualization. All of these factors have given rise to the most recent direction of Explainable AI (XAI).
Augmenting AI with the facets of human centricity becomes indispensable. It is desirable that the models of AI are transparent so that the results being produced are easily interpretable and explainable. There have been a number of studies identifying opaque constructs of artificial neural networks (including deep learning) and stressing centrality of ways of bringing the aspect of transparency to the developed constructs. To make the results interpretable and deliver the required facet of explainability, one may argue that the findings have to be delivered at a certain level of abstraction (casting them in some general perspective)—subsequently information granularity and information granules play here a pivotal role. Likewise, the explanation mechanisms could be inherently associated with the logic fabric of the constructs, which facilitate the realization of interpretation and explanation processes. These two outstanding features help carry out a thorough risk analysis associated with actionable actions based on conclusions delivered by the AI system.
The volume provides the readers with a comprehensive and up-to-date treatise on the studies at the junction of the area of XAI and Granular Computing. The chapters contributed by active researchers and practitioners exhibit substantial diversity naturally reflecting the breadth of the area itself. The methodology, advanced algorithms and case studies and applications are covered. XAI for processing mining, visual analytics, knowledge, learning, and interpretation are among the highly representative trends in the area. The applications to text classification, image processing prediction covered by several chapters are a tangible testimony to the recent advancements of AI.
Скачать Interpretable Artificial Intelligence: A Perspective of Granular Computing