| 1. |
- Bacauskiene, Marija, et al.
(författare)
-
Selecting salient features for classification based on neural network committees
- 2004
-
Ingår i: Pattern Recognition Letters. - Amsterdam : Elsevier Science. - 0167-8655 .- 1872-7344. ; 25:16, s. 1879-1891
-
Tidskriftsartikel (refereegranskat)abstract
- Aggregating outputs of multiple classifiers into a committee decision is one of the most important techniques for improving classification accuracy. The issue of selecting an optimal subset of relevant features plays also an important role in successful design of a pattern recognition system. In this paper, we present a neural network based approach for identifying salient features for classification in neural network committees. Feature selection is based on two criteria, namely the reaction of the cross-validation data set classification error due to the removal of the individual features and the diversity of neural networks comprising the committee. The algorithm developed removed a large number of features from the original data sets without reducing the classification accuracy of the committees. The accuracy of the committees utilizing the reduced feature sets was higher than those exploiting all the original features.
|
|
| 2. |
- Bacauskiene, Marija, et al.
(författare)
-
The Evidence Theory Based Post-Processing of Colour Images
- 2004
-
Ingår i: Informatica (Vilnius). - Vilnius : Institute of Mathematics and Cybernetics, Lithuanian Academy of Sciences. - 0868-4952 .- 1822-8844. ; 15:3, s. 315-328
-
Tidskriftsartikel (refereegranskat)abstract
- The problem of post-processing of a classified image is addressed from the point of view of the Dempster-Shafer theory of evidence. Each neighbour of a pixel being analyzed is considered as an item of evidence supporting particular hypotheses regarding the class label of that pixel. The strength of support is defined as a function of the degree of uncertainty in class label of the neighbour, and the distance between the neighbour and the pixel being considered. A post-processing window defines the neighbours. Basic belief masses are obtained for each of the neighbours and aggregated according to the rule of orthogonal sum. The final label of the pixel is chosen according to the maximum of the belief function.
|
|
| 3. |
- Bergman, Linda, et al.
(författare)
-
Intelligent Monitoring of the Offset Printing Process
- 2004
-
Ingår i: Neural Networks and Computational Intelligence - Proceedings. - : ACTA Press. ; , s. 173-178, s. 173-178
-
Konferensbidrag (refereegranskat)abstract
- In this paper, we present a neural networks and image analysis based approach to assessing colour deviations in an offset printing process from direct measurements on halftone multicoloured pictures--there are no measuring areas printed solely to assess the deviations. A committee of neural networks is trained to assess the ink proportions in a small image area. From only one measurement the trained committee is capable of estimating the actual amount of printing inks dispersed on paper in the measuring area. To match the measured image area of the printed picture with the corresponding area of the original image, when comparing the actual ink proportions with the targeted ones, properties of the 2-D Fourier transform are exploited.
|
|
| 4. |
- Bergman, Lars, et al.
(författare)
-
Modelling and Control of the Web-Fed Offset Newspaper Printing Press
- 2003
-
Ingår i: Proceedings of the Technical Association of the Graphic Arts, TAGA. - : Technical Association of the Graphic (TAGA). ; , s. 27-29
-
Konferensbidrag (refereegranskat)abstract
- We present an approach to modelling and controlling the web-fed offset printing process. An image processing and artificial neural networks based device is used to measure the printing process output - the observable variables. The observable variables are measured on halftone areas and integrate information about both ink densities and dot sizes. From only one measurement the device is capable of estimating the actual relative amount of each cyan, magenta, yellow, and black ink dispersed on paper in the measuring area. We build and test linear and non-linear printing press models using the measured variables andother parameters characterising the press. The observable variables measured and the press model developed are then further used by a control unit for generating control signals - signals for controlling the ink keys - to compensate for colour deviation. The experimental investigations performed have shown that the non-linear model developed is accurate enough to be used in a control loop for controlling the printing process. The control accuracy - the tracking accuracy of the desired ink level - obtained from the controller was higher than that observed when controlling the press by the operator.
|
|
| 5. |
- Stasiunas, Antanas, et al.
(författare)
-
A non-linear circuit for simulating OHC of the cochlea
- 2003
-
Ingår i: Medical Engineering and Physics. - London : Elsevier. - 1350-4533 .- 1873-4030. ; 25:7, s. 591-601
-
Tidskriftsartikel (refereegranskat)abstract
- In the present paper, referring to known characteristics of the outer hair cells functioning in the cochlea of the inner ear, a functional model of the outer hair cells is constructed. It consists of a linear feed-forward circuit and a non-linear positive feedback circuit. The feed-forward circuit reflects the contribution of local basilar and tectorial membrane areas and passive outer hair cells’ physical parameters to the forming of low-selectivity resonance characteristics. The non-linear positive feedback circuit reflects the non-linear outer hair cell signal transduction processes and the active role of efferents from the medial superior olive in altering circuit sensitivity and selectivity.Referring to an analytical description of the circuit model and computer simulation results, an explanation is given over the biological meaning of the outer hair cells’ non-linearities in signal transduction processes and the role of the non-linearities in achieving the following: signal compression, the dependency of circuit sensitivity and frequency selectivity upon the input signal amplitude, the compatibility of high-frequency selectivity and short transient response of the biological filtering circuits.
|
|
| 6. |
- Verikas, Antanas, et al.
(författare)
-
An intelligent system for tuning magnetic field of a cathode ray tube deflection yoke
- 2003
-
Ingår i: Knowledge-Based Systems. - Amsterdam : Elsevier Science. - 0950-7051 .- 1872-7409. ; 16:3, s. 161-164
-
Tidskriftsartikel (refereegranskat)abstract
- This short communication concerns identification of the number of magnetic correction shunts and their positions for deflection yoke tuning to correct the misconvergence of colours of a cathode ray tube. The misconvergence of colours is characterised by the distances measured between the traces of red and blue beams. The method proposed consists of two phases, namely, learning and optimisation. In the learning phase, the radial basis function neural network is trained to learn a mapping: correction shunt position→changes in misconvergence. In the optimisation phase, the trained neural network is used to predict changes in misconvergence depending on a correction shunt position. An optimisation procedure based on the predictions returned by the neural net is then executed in order to find the minimal number of correction shunts needed and their positions. During the experimental investigations, 98% of the deflection yokes analysed have been tuned successfully using the technique proposed.
|
|
| 7. |
- Verikas, Antanas, 1951-, et al.
(författare)
-
Colour speck counter for assessing the dirt level in secondary fibre pulps
- 2003
-
Ingår i: Journal of Pulp and Paper Science (JPPS). - Montreal : Pulp and Paper Technical Association of Canada. - 0826-6220. ; 29:7, s. 220-224
-
Tidskriftsartikel (refereegranskat)abstract
- Speck count is increasingly used as a parameter to assess the quality of secondary fibre pulps. The resolution of most of the commercial image analysis systems is too low for detecting small specks. Therefore, small specks are not taken into consideration when using conventional image analysis systems to assess pulp quality. We have recently developed a colour speck counter which can detect specks ranging in size from ∼5 to 300 μm. In this paper, we present the results of experimental investigations related to the use of the speck counter to assess the dirt level in secondary fibre pulps. We assume an exponential speck size distribution and advocate the idea of using the scale parameter λ of the distribution to characterize the size content of a set of specks detected. Experimental investigations performed have shown that the scale parameter, together with the expected speck area and the speck number, can be used to characterize and rank secondary fibre pulps according to dirt level and the dirt-size distribution.
|
|
| 8. |
- Verikas, Antanas, 1951-, et al.
(författare)
-
Combining neural networks, fuzzy sets, and evidence theory based approaches for analysing colour images
- 2000
-
Ingår i: IJCNN 2000. - Los Alamitos : IEEE Computer Society. - 9780769506197 - 0769506194 - 0780365410 - 9780780365414 - 0769506216 - 9780769506210 ; , s. 297-302
-
Konferensbidrag (refereegranskat)abstract
- This paper presents an approach to determining colours of specks in an image taken from a pulp sample. The task is solved through colour classification by an artificial neural network. The network is trained using possibilistic target values. The problem of post-processing of a pixelwise-classified image is addressed from the point of view of the Dempster-Shafer theory of evidence. Each neighbour of a pixel being analysed is considered as an item of evidence supporting particular hypotheses regarding the class label of that pixel. The experiments performed have shown that the colour classification results correspond well with the human perception of colours of the specks.
|
|
| 9. |
- Verikas, Antanas, 1951-, et al.
(författare)
-
Combining neural networks, fuzzy sets, and the evidence theory based techniques for detecting colour specks
- 2001
-
Ingår i: Journal of Intelligent & Fuzzy Systems. - Amsterdam : IOS Press. - 1064-1246 .- 1875-8967. ; 10:2, s. 117-130
-
Tidskriftsartikel (refereegranskat)abstract
- An approach to detecting colour specks in an image taken from a pulp sample of recycled paper is presented. The task is solved through pixel-wise colour classification by an artificial neural network and post-processing based on the evidence theory. The network is trained using possibilistic target values, which are determined through a self-organising process in a 2D and 1D map of chromaticity and lightness, respectively. The problem of post-processing of a pixelwise-classified image is addressed from the point of view of the Dempster-Shafer theory of evidence. Each neighbour of a pixel being analysed is considered as an item of evidence supporting particular hypotheses regarding the class label of that pixel. The strength of support is defined as a function of the degree of uncertainty in class label of the neighbour, and the distance between the neighbour and the pixel being considered. The experiments performed have shown that the colour classification results correspond well with the human perception of colours of the specks.
|
|
| 10. |
- Verikas, Antanas, et al.
(författare)
-
Feature Selection with Neural Networks
- 2002
-
Ingår i: Pattern Recognition Letters. - Amsterdam : Elsevier. - 0167-8655 .- 1872-7344. ; 23:11, s. 1323-1335
-
Tidskriftsartikel (refereegranskat)abstract
- We present a neural network based approach for identifying salient features for classification in feedforward neural networks. Our approach involves neural network training with an augmented cross-entropy error function. The augmented error function forces the neural network to keep low derivatives of the transfer functions of neurons when learning a classification task. Such an approach reduces output sensitivity to the input changes. Feature selection is based on the reaction of the cross-validation data set classification error due to the removal of the individual features. We demonstrate the usefulness of the proposed approach on one artificial and three real-world classification problems. We compared the approach with five other feature selection methods, each of which banks on a different concept. The algorithm developed outperformed the other methods by achieving higher classification accuracy on all the problems tested.
|
|
