Data dimensionality reduction with application to simplifying RBF network structure and improving classification performance

@article{FU2003DataDR,
  title={Data dimensionality reduction with application to simplifying RBF network structure and improving classification performance},
  author={Xiuju FU and Lipo Wang},
  journal={IEEE transactions on systems, man, and cybernetics. Part B, Cybernetics : a publication of the IEEE Systems, Man, and Cybernetics Society},
  year={2003},
  volume={33 3},
  pages={
          399-409
        }
}
  • Xiuju FU, Lipo Wang
  • Published 1 June 2003
  • Computer Science
  • IEEE transactions on systems, man, and cybernetics. Part B, Cybernetics : a publication of the IEEE Systems, Man, and Cybernetics Society
For high dimensional data, if no preprocessing is carried out before inputting patterns to classifiers, the computation required may be too heavy. For example, the number of hidden units of a radial basis function (RBF) neural network can be too large. This is not suitable for some practical applications due to speed and memory constraints. In many cases, some attributes are not relevant to concepts in the data at all. In this paper, we propose a novel separability-correlation measure (SCM) to… 
View on PubMed
dr.ntu.edu.sg
294 Citations
Highly Influential Citations
15
Background Citations
94
Methods Citations
83
Results Citations
2

294 Citations

Citation Type

Citation Type

Data dimensionality reduction with application to improving classification performance and explaining concepts of data sets
TLDR
A rule-extraction method based on a novel separability-correlation measure (SCM) ranking the importance of attributes that leads to smaller rule sets with higher accuracies compared with other methods.
Study of Dimensionality Reduction Techniques for Effective Investment Portfolio Data Management
TLDR
It has been observed that cfsSubset evaluator with partition membership as a pre-processing technique and random forest as classification algorithm performs reasonably better in terms of accuracy.
Feature selection using localized generalization error for supervised classification problems using RBFNN
A Genetic Approach to Data Dimensionality Reduction Using a Special Initial Population
TLDR
It is shown that a wise selection of the initial population improves the performance of the DDR-GA considerably and a method to implement this approach is introduced and results indicate that the approach provides more accurate results compared to the purely random one in a reasonable amount of time.
A New Approach to Division of Attribute Space for SVR Based Classification Rule Extraction
TLDR
A new measure for determining the importance level of the attributes based on the trained SVR (Support vector re-gression) classifiers is proposed and a new approach for the division to continuous attribute space based on support vectors is given.
A New Approach to Attribute Importance Ranking for Constructing Classification Rules Based on SVR
TLDR
The differential characteristic of SVR (Support vector regression) is discussed and a new measure for determining the importance level of the attributes based on the trained SVR classifiers is proposed, which can improve the validity of the extracted rules remarkably compared to other rule extracting approaches.
Assessment of prediction ability for reduced probabilistic neural network in data classification problems
TLDR
This paper shows that the algorithm based on k-means clustering is a better PNN structure reduction procedure and is much less time-consuming than the standard tenfold cross-validation procedure.
Efficient construction of sparse radial basis function neural networks using L1-regularization
Improving multiclass pattern recognition with a co-evolutionary RBFNN
Data Dimensionality Reduction Based on Derivative Characteristics of Trained Support Vector Regression
TLDR
A new measure for determining the importance level of the attributes based on the trained support vector regression (SVR) and its derivative characteristics is proposed and can improve efficiency and effectiveness significantly compared with other data dimensionality reduction approaches.
...
1
2
3
4
5
...

References

SHOWING 1-10 OF 24 REFERENCES
Dimensionality reduction of unsupervised data
  • M. Dash, Hua Liu, J. Yao
  • Computer Science
    Proceedings Ninth IEEE International Conference on Tools with Artificial Intelligence
  • 1997
TLDR
This paper proposes an entropy measure for ranking features, and conducts extensive experiments to show that the method is able to find the important features and compares well with a similar feature ranking method that requires class information unlike this method.
Efficient Training Of Rbf Networks For Classification
  • I. Nabney
  • Computer Science
    Int. J. Neural Syst.
  • 2004
TLDR
This work shows how RBFs with logistic and softmax outputs can be trained efficiently using the Fisher scoring algorithm, and compares this approach with standard non-linear optimisation algorithms on a number of datasets.
Using mutual information for selecting features in supervised neural net learning
  • R. Battiti
  • Computer Science
    IEEE Trans. Neural Networks
  • 1994
This paper investigates the application of the mutual information criterion to evaluate a set of candidate features and to select an informative subset to be used as input data for a neural network…
Enhanced Fisher linear discriminant models for face recognition
  • C. Liu, H. Wechsler
  • Computer Science
    Proceedings. Fourteenth International Conference on Pattern Recognition (Cat. No.98EX170)
  • 1998
TLDR
Two enhanced Fisher linear discriminant models (EFM) are introduced in order to improve the generalization ability of the standard FLD based classifiers such as Fisherfaces and Experimental data shows that the EFM models outperform the standardFLD based methods.
A new method for initializing radial basis function classifiers
  • T. Kaylani, S. DasGupta
  • Computer Science
    Proceedings of IEEE International Conference on Systems, Man and Cybernetics
  • 1994
TLDR
A new approach for the selection of RBF kernel centers and their effective widths is introduced and validated using a speaker-independent vowel recognition problem.
Estimating Attributes: Analysis and Extensions of RELIEF
TLDR
In the context of machine learning from examples this paper deals with the problem of estimating the quality of attributes with and without dependencies among them and is analysed and extended to deal with noisy, incomplete, and multi-class data sets.
RBF and CBF neural network learning procedures
TLDR
To achieve fast convergence, the learning algorithms were investigated to find fast and efficient procedures to automatically extract fuzzy rules and membership functions from high dimensional data.
A neural-network learning theory and a polynomial time RBF algorithm
TLDR
A new learning theory (a set of principles for brain-like learning) and a corresponding algorithm for the neural-network field are presented and computational results are provided for the well known Mackey-Glass chaotic time series problem, the logistic map prediction problem, various neuro-control problems, and several time series forecasting problems.
Approximate radial basis function neural networks (RBFNN) to learn smooth relations from noisy data
  • P. Maffezzoni, P. Gubian
  • Computer Science
    Proceedings of 1994 37th Midwest Symposium on Circuits and Systems
  • 1994
In this paper a novel RBFNN scheme is presented introducing the idea of strongly delocalized neural receptive fields. Based on delocalization, a robust deterministic annealing procedure is proposed…
An algorithm to generate radial basis function (RBF)-like nets for classification problems
...
1
2
3
...