Incremental Updating Multigranulation Approximations with Matrix Representation Under Coarsening Decision Attribute

Abstract

Incremental learning is a computing method for dynamic systems. In this paper, we study the coarsening of decision attribute values in multi-granularity and define some matrixes and multi granularity fusion method based on matrix representation. In addition, we design an algorithm for calculating approximations. Furthermore, we describe the coarsening method of decision attributes, and update the multi-granularity probability approximations by updating matrixes and propose related algorithms. Compared with non-incremental algorithm, incremental algorithm can effectively accelerate knowledge acquisition.