Семинар ММИТ — Granular Clustering: Augmenting Principles, Realizing Symbolic-Granular Descriptions and Application-Oriented Implications
В четверг, 26 апреля 2018 года, состоится очередное заседание семинара "Математические модели информационных технологий" департамента анализа данных и искусственного интеллекта и МНУЛ "Интеллектуальные системы и структурный анализ" под руководством С.О. Кузнецова.
Место проведения: Кочновский проезд, 3. ауд. 622, 16:40
Title: Granular Clustering: Augmenting Principles, Realizing Symbolic-Granular Descriptions and Application-Oriented Implications
Speaker: Dr. Witold Pedrycz, Professor and Chair Canada Research Chair IEEE Fellow Professional Engineer Department of Electrical and. Computer Engineering, University of Alberta, Edmonton, Canada
Abstract: Clustering has been for decades a focal point of studies quite often researched in relation with modeling, pattern classification, and data analysis. With the advent of data analytics bringing a suite of new problems, clustering has been subjected to a visible paradigm shift. Granular clustering, the term being recently used, has emphasized the role of clustering regarded as a sound vehicle to construct information granules – entities aimed at the building abstract yet flexible and adjustable views at data, facilitating processing of masses of data and subsequently constructing interpretable models.
The term granular clustering can be sought from the two general points of view; in this talk those perspectives are carefully analyzed along with a formulation of far reaching ramifications. The first general view is concerned with the formation of information granules completed on a basis of predominantly numeric (non-granular) data. The alternative view stresses the clustering of granular data themselves. The hybrid architectures of these views are also investigated.
While the results of clustering algorithms are conveyed through numeric constructs (say, prototypes and partition matrices, etc.), we discuss here an attractive alternative of symbolic (qualitative) characterization of information granules (clusters), which supports higher levels of interpretability and offers insights into aspects of stability of structural findings.
In the setting of data analytics, there are several well-articulated and emerging challenges. Considering objective function-based clustering, these techniques return a small number of numeric representatives (prototypes) of big data. This triggers a question as to the representation capabilities of the prototypes. A certain line of research is to augment the numeric prototypes produced by their granular generalizations (viz. granular prototypes) and optimize their abilities to capture the essence of the data. We discuss a direction of research aimed at building optimal granular prototypes and their characterization. It is shown that some clustering techniques exhibiting a great deal of flexibility (such as e.g., DBSCAN or hierarchical clustering) still require a concise characterization of the comprehensive results coming in the form of granular prototypes. An impact on ensuing modeling (viz. modeling exploiting granular data) is discussed.
Clustering techniques are commonly concerned with the formation of direction-free (relational) constructs such as those being used in association (linkage) analysis. The accommodation of the aspect of directionality (required to cope with in various modeling tasks) entails another wave of pursuits that are referred to as direction-sensitive clustering.
Приглашаются все желающие. Для студентов, аспирантов, преподавателей и сотрудников Высшей школы экономики - вход свободный. При необходимости заказа пропуска в здание НИУ ВШЭ просьба сообщить по e-mail: firstname.lastname@example.org