Title :
Automatic input space partitioning for hierarchical fuzzy systems
Author_Institution :
Bavarian Res. Center for Knowledge-Based Syst., Erlangen, Germany
Abstract :
The paper is based on a new learning algorithm for Hierarchical Fuzzy Associative Memories (HIFAM) recently proposed by the author. It has been demonstrated that the proposed method is suited for classification problems as well as for regression tasks and that it compares well to existing machine learning techniques. The paper investigates an extension to the HIFAM method that allows the simultaneous creation of rules and fuzzy sets. Several results of the approach on commonly used benchmark data sets are given and compared to the results of the original algorithm
Keywords :
content-addressable storage; fuzzy set theory; fuzzy systems; learning (artificial intelligence); pattern classification; HIFAM method; Hierarchical Fuzzy Associative Memories; automatic input space partitioning; benchmark data sets; classification problems; fuzzy sets; hierarchical fuzzy systems; learning algorithm; machine learning techniques; regression tasks; simultaneous creation; Associative memory; Binary trees; Fuzzy sets; Fuzzy systems; Knowledge based systems; Machine learning; Machine learning algorithms; Partitioning algorithms; Quantization; Training data;
Conference_Titel :
Fuzzy Information Processing Society - NAFIPS, 1998 Conference of the North American
Conference_Location :
Pensacola Beach, FL
Print_ISBN :
0-7803-4453-7
DOI :
10.1109/NAFIPS.1998.715578
