Artificial Intelligence and Granular Computing:
New Directions in System Design and Analysis
Witold Pedrycz Abstract
A number of recent advancements in Artificial Intelligence (AI) fall under the umbrella of explainable AI (XAI). We advocate that in the realization of these pursuits, information granules and Granular Computing play a significant role. First, it is shown that information granularity is of paramount relevance in building linkages between real-world data and symbols commonly encountered in AI processing. Second, we stress that a suitable level of abstraction (realized through selecting a required level of information granularity) becomes essential to support user- oriented framework of design of AI artifacts. Third, it is articulated that a logic fabric supporting interpretation faculties and promoting the modularity becomes instrumental to facilitate the transparency of the established constructs.
In the scenarios outlined above, central to all developments is a process of formation of information granules and a prudent characterization of their quality. We discuss a comprehensive approach to the development of information granules being completed by means of clustering and the principle of justifiable granularity. Here various construction scenarios are discussed with regard to the accommodation of available domain knowledge. We look at the generative and discriminative aspects of information granules.
Highly-dimensional data come with a concentration effect. Its presence calls for a design shift by moving towards forming qualitative descriptors (symbols) and their summarization.We discuss ideas, architectures and algorithms of (i) logic-oriented factorization and (ii) low-dimensional rule-based models and their aggregation and evaluation.
The talk is made self-contained and covers required essentials of AI and Granular Computing.