Information theoretic subset selection for neural network models

Sridhar, Dasaratha V.; Bartlett, E.B.; Seagrave, Richard C.

doi:10.1016/s0098-1354(97)00227-5

Cited by 53 publications

(34 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Extensive tests including novel combinations of ranking and selection methods will be published elsewhere. A very simple dataset used by Shridhar et al [28] contains 12 patterns with 4 features (Tab. 2).…”

Section: Illustrative Results On Synthetic Datamentioning

confidence: 99%

“…For the Shridhar dataset almost all ranking and redundancy shifting methods worked correctly. If the relevancy threshold is set to 0.05, features marked in bold (irrelevant) Shridhar Corral Gauss8 dataset [28] dataset [29] dataset [21,24] Table 3. Ordering of features after feature selection for 3 synthetic datasets.…”

Section: Illustrative Results On Synthetic Datamentioning

confidence: 99%

See 1 more Smart Citation

Infosel++: Information Based Feature Selection C++ Library

Kachel¹,

Biesiada²,

Blachnik³

et al. 2010

Artificial Intelligence and Soft Computing

View full text Add to dashboard Cite

Abstract. A large package of algorithms for feature ranking and selection has been developed. Infosel++, Information Based Feature Selection C++ Library, is a collection of classes and utilities based on probability estimation that can help developers of machine learning methods in rapid interfacing of feature selection algorithms, aid users in selecting an appropriate algorithm for a given task (embed feature selection in machine learning task), and aid researchers in developing new algorithms, especially hybrid algorithms for feature selection. A few examples of such possibilities are presented.

show abstract

Section: Illustrative Results On Synthetic Datamentioning

confidence: 99%

Section: Illustrative Results On Synthetic Datamentioning

confidence: 99%

Infosel++: Information Based Feature Selection C++ Library

Kachel¹,

Biesiada²,

Blachnik³

et al. 2010

Artificial Intelligence and Soft Computing

View full text Add to dashboard Cite

show abstract

“…We selected this software for our pattern recognition system because of its extensive and sophisticated proven capabilities at solving difficult data mining problems [3], [26]. The set of weight values and number of nodes for an ANN is called the ANN architecture.…”

Section: Pattern Recognition Systemmentioning

confidence: 99%

An artificial neural network for wavelet steganalysis

2005

View full text Add to dashboard Cite

Hiding messages in image data, called steganography, is used for both legal and illicit purposes. The detection of hidden messages in image data stored on websites and computers, called steganalysis, is of prime importance to cyber forensics personnel. Automating the detection of hidden messages is a requirement, since the shear amount of image data stored on computers or websites makes it impossible for a person to investigate each image separately. This paper describes research on a prototype software system that automatically classifies an image as having hidden information or not, using a sophisticated artificial neural network (ANN) system. An ANN software package, the ISU ACL NetWorks Toolkit, is trained on a selection of image features that distinguish between stego and nonstego images. The novelty of this ANN is that it is a blind classifier that gives more accurate results than previous systems. It can detect messages hidden using a variety of different types of embedding algorithms. A Graphical User Interface (GUI) combines the ANN, feature selection, and embedding algorithms into a prototype software package that is not currently available to the cyber forensics community. ABSTRACTHiding messages in image data, called steganography, is used for both legal and illicit purposes. The detection of hidden messages in image data stored on websites and computers, called steganalysis, is of prime importance to cyber forensics personnel. Automating the detection of hidden messages is a requirement, since the shear amount of image data stored on computers or websites makes it impossible for a person to investigate each image separately. This paper describes research on a prototype software system that automatically classifies an image as having hidden information or not, using a sophisticated artificial neural network (ANN) system. An ANN software package, the ISU ACL NetWorks Toolkit, is trained on a selection of image features that distinguish between stego and nonstego images. The novelty of this ANN is that it is a blind classifier that gives more accurate results than previous systems. It can detect messages hidden using a variety of different types of embedding algorithms. A Graphical User Interface (GUI) combines the ANN, feature selection, and embedding algorithms into a prototype software package that is not currently available to the cyber forensics community.

show abstract

“…Another ratio IGn(X j ) = IG(X j )/I(C), called also "an asymmetric dependency coefficient" is advocated in [6]. Mutual information between feature f and classes:…”

Section: Information Theory and Other Filtersmentioning

confidence: 99%

“…Due to the space restrictions we report here only results obtained with information gain (IGn) ranking [6] and Battiti selection method (BA) [7] on two datasets [10]: Monk-1 artificial data and hypothyroid problems. Monk-1 data has only 6 features of which 5, 1, 2 are used to form a rule determining the class.…”

Section: Numerical Experimentsmentioning

confidence: 99%

Feature selection based on information theory, consistency and separability indices

Duch

Grąbczewski

Winiarski

et al.

Proceedings of the 9th International Conference on Neural Information Processing, 2002. ICONIP '02.

View full text Add to dashboard Cite

Two new feature selection methods are introduced, the first based on separability criterion, the second on consistency index that includes interactions between the selected subsets of features. Comparison of accuracy was made against information-theory based selection methods on several datasets training neurofuzzy and nearest neighbor methods on various subsets of selected features. Methods based on separability seem to be most promising.

show abstract

Information theoretic subset selection for neural network models

Cited by 53 publications

References 16 publications

Infosel++: Information Based Feature Selection C++ Library

Infosel++: Information Based Feature Selection C++ Library

An artificial neural network for wavelet steganalysis

Feature selection based on information theory, consistency and separability indices

Contact Info

Product

Resources

About