Rough Sets and Granular Computing

From Nafips2009

Jump to: navigation, search

Rough set theory is a computational model of approximate reasoning proposed by Z. Pawlak in 1982. The main idea is that human observations may not be able to measure or describe accurately concepts given by the nature. Pawlak uses two approximate measurements or descriptions: a set of sufficient conditions, such as a set of manifestations called lower approximation, and a set of necessary conditions, such as a collection of all possible symptoms observed in patients of flu, called upper approximation.

Although rough sets have been studied independently, L.A. Zadeh gave a new insight into this field by observing rough set theory is a crisp-based theory of granular computing, which is a label proposed by T. Y. Lin (and L. A. Zadeh) to name the notion rooted deeply in fuzzy theory:Zadeh (1979) stated "information graulearity . . . to decomposition and partition-- in the theory of automata and system science . . . to bounded uncertainties -- in optimal control. . ." To promote the notion, Lin (and Zadeh) started a special interest group on granular computing (SIGGrC) in BISC (Berkeley Institute of Soft Computing) during his sabbatical leave to Berkeley (1996-97).

Rough sets apply fundamentals of set classification to database analysis (equivalence classes are information granules), which is a crisp-set based granular computing. On the other hand, in granular computing, new theory and methods are proposed involving not only rough sets but also fuzzy sets, probability (e.g., granular probability by G. Klir) multisets, neural networks, belief networks, modal logic, rule induction method.

This special session gives a special opportunity for researchers on rough sets and granular computing to exchange their ideas.

Personal tools