Paper Info
Title
Incorporating fuzzy concepts along with dynamic tunneling for fast and robust training of multilayer perceptrons
Abstract
This paper proposes a new computational technique for training of fuzzified multilayer perceptrons (MLP), which is based on the fuzzification of input and output along with recently proposed dynamic tunneling technique in MLP. The fuzzified MLP generates fuzzy output from fuzzy input whose connection weights are real numbers. This type of modeling allows realistic representation of natural/real life data which is akin to human reasoning process. The proposed technique is applied to problems like; approximate realization of non-linear mapping of fuzzy numbers, classification of fuzzy patterns, and approximate realization of fuzzy if then rules. The simulation results are provided for eight test examples to demonstrate the learning capabilities of the proposed method. The generalization property of MLP in learning using the above technique is also discussed. Lastly, the proposed technique is applied to some well-known benchmark problems to assess the classification ability.
Year
DOI
Venue
2003
10.1016/S0925-2312(02)00570-2
Neurocomputing
Keywords
Field
DocType
Lipschitz condition,Dynamic tunneling,Fuzzy MLP
Neuro-fuzzy,Defuzzification,Fuzzy classification,Fuzzy set operations,Computer science,Fuzzy logic,Fuzzy set,Artificial intelligence,Fuzzy associative matrix,Fuzzy number,Machine learning
Journal
Volume
ISSN
Citations 
50
0925-2312
4
PageRank 
References 
Authors
0.44
8
2
Name
Order
Citations
PageRank
Pinaki Roy Chowdhury161.66
K. K. Shukla2392.13