Name
Abstract | ||
|---|---|---|
Recently, many works have begun to explore the possibility of using samples out of the training set to improve their results. One of these approaches, Universum, introduced by Vapnik had already been used in combination with several classifiers to increase their accuracy. In this paper, we present a novel approach for identifying the consequent parameters of the Takagi-Sugeno Fuzzy Model, UFuzzy. It is based on the idea of the Universum set, which acts to regularize the optimization problem. It also helps with the introduction of prior knowledge to the tasks performed by the model. In addition, we explore the influence of the Universum set on identifying the structure of fuzzy rules using the c-means clustering algorithm. We evaluated our approach on several generated and real-world classification datasets and it shows promising results in comparison to the baseline methods, which do not use the Universum set. |
Year | DOI | Venue |
|---|---|---|
2017 | 10.1016/j.asoc.2017.05.018 | Applied Soft Computing |
Keywords | DocType | Volume |
TS-Fuzzy models,Universum learning,Unlabeled data,Regularization,Semi-supervised learning | Journal | 59 |
ISSN | Citations | PageRank |
1568-4946 | 1 | 0.35 |
References | Authors | |
24 | 3 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Lukas Tencer | 1 | 26 | 3.78 |
| Marta Reznáková | 2 | 20 | 4.02 |
| Mohamed Cheriet | 3 | 2047 | 238.58 |