Abstract | ||
---|---|---|
The curse of dimensionality is one of the most severe problems for the application of RBF networks: The number of RBF nodes and the number of required training examples grow exponentially with the intrinsic dimensionality of the input space. One way to address this problem is the application of feature selection as a data pre-processing step.Here, we introduce a two-step feature selection process: Firstly, potentially optimal feature subsets are pre-selected by the feature selection technique EUBAFES-DW. Secondly, RBF networks are applied to determine which of the subsets lead to least network complexity and best classification accuracy.We give a detailed description of EUBAFES-DW and show the capability of the two-step approach to improve RBF networks by its application to a number of artificial and real world data sets. |
Year | Venue | Field |
---|---|---|
1997 | PROGRESS IN CONNECTIONIST-BASED INFORMATION SYSTEMS, VOLS 1 AND 2 | Pattern recognition,Feature selection,Computer science,Artificial intelligence,Machine learning |
DocType | Citations | PageRank |
Conference | 0 | 0.34 |
References | Authors | |
0 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Matthias Scherf | 1 | 0 | 0.34 |
Wilfried Brauer | 2 | 969 | 299.36 |