2014 |
Homenda, Wladyslaw; Luckner, Marcin Pattern recognition with rejection: Application to handwritten digits Inproceedings 2014 4th World Congress on Information and Communication Technologies (WICT 2014), pp. 326–331, IEEE, 2014, ISBN: 978-1-4799-8115-1. Abstract | Links | BibTeX | Tags: Accuracy, Classification with rejection, Handwriting recognition, native and foreign elements, pattern recognition with rejection, Standards, Support Vector Machines, Testing, Text recognition @inproceedings{Homenda2014a, title = {Pattern recognition with rejection: Application to handwritten digits}, author = {Wladyslaw Homenda and Marcin Luckner}, url = {http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=7077288}, doi = {10.1109/WICT.2014.7077288}, isbn = {978-1-4799-8115-1}, year = {2014}, date = {2014-12-01}, booktitle = {2014 4th World Congress on Information and Communication Technologies (WICT 2014)}, pages = {326--331}, publisher = {IEEE}, abstract = {The paper considers rejecting option in pattern recognition problem. Studied are native and foreign elements in a multi-class pattern recognition. Native elements are those included in recognized classes, they are known at the stage of classifier design. Foreign elements do not belong to recognized classes. Usually foreign elements are not known when classifier is designed. If foreign elements are classified to recognized classes, recognition quality is deteriorated. So then, they are classified to native classes, if they are not rejected. In such the case, recognition quality is deteriorated. Therefore, they should be rejected by a classifier, i.e. not classified to any class. Several attempts to rejection of foreign elements are investigated in this study.}, keywords = {Accuracy, Classification with rejection, Handwriting recognition, native and foreign elements, pattern recognition with rejection, Standards, Support Vector Machines, Testing, Text recognition}, pubstate = {published}, tppubtype = {inproceedings} } The paper considers rejecting option in pattern recognition problem. Studied are native and foreign elements in a multi-class pattern recognition. Native elements are those included in recognized classes, they are known at the stage of classifier design. Foreign elements do not belong to recognized classes. Usually foreign elements are not known when classifier is designed. If foreign elements are classified to recognized classes, recognition quality is deteriorated. So then, they are classified to native classes, if they are not rejected. In such the case, recognition quality is deteriorated. Therefore, they should be rejected by a classifier, i.e. not classified to any class. Several attempts to rejection of foreign elements are investigated in this study. |
Homenda, Wladyslaw; Luckner, Marcin Pattern recognition with rejection: Application to handwritten digits Inproceedings 2014 4th World Congress on Information and Communication Technologies (WICT 2014), pp. 326–331, IEEE, 2014, ISBN: 978-1-4799-8115-1. Abstract | Links | BibTeX | Tags: Accuracy, Classification with rejection, Handwriting recognition, native and foreign elements, pattern recognition with rejection, Standards, Support Vector Machines, Testing, Text recognition @inproceedings{Homenda2014ab, title = {Pattern recognition with rejection: Application to handwritten digits}, author = {Wladyslaw Homenda and Marcin Luckner}, url = {http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=7077288}, doi = {10.1109/WICT.2014.7077288}, isbn = {978-1-4799-8115-1}, year = {2014}, date = {2014-12-01}, booktitle = {2014 4th World Congress on Information and Communication Technologies (WICT 2014)}, pages = {326--331}, publisher = {IEEE}, abstract = {The paper considers rejecting option in pattern recognition problem. Studied are native and foreign elements in a multi-class pattern recognition. Native elements are those included in recognized classes, they are known at the stage of classifier design. Foreign elements do not belong to recognized classes. Usually foreign elements are not known when classifier is designed. If foreign elements are classified to recognized classes, recognition quality is deteriorated. So then, they are classified to native classes, if they are not rejected. In such the case, recognition quality is deteriorated. Therefore, they should be rejected by a classifier, i.e. not classified to any class. Several attempts to rejection of foreign elements are investigated in this study.}, keywords = {Accuracy, Classification with rejection, Handwriting recognition, native and foreign elements, pattern recognition with rejection, Standards, Support Vector Machines, Testing, Text recognition}, pubstate = {published}, tppubtype = {inproceedings} } The paper considers rejecting option in pattern recognition problem. Studied are native and foreign elements in a multi-class pattern recognition. Native elements are those included in recognized classes, they are known at the stage of classifier design. Foreign elements do not belong to recognized classes. Usually foreign elements are not known when classifier is designed. If foreign elements are classified to recognized classes, recognition quality is deteriorated. So then, they are classified to native classes, if they are not rejected. In such the case, recognition quality is deteriorated. Therefore, they should be rejected by a classifier, i.e. not classified to any class. Several attempts to rejection of foreign elements are investigated in this study. |
Luckner, Marcin Global and local rejection option in multi-classification task Inproceedings Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), pp. 483–490, 2014, ISSN: 16113349. Abstract | Links | BibTeX | Tags: Graph Ensemble, Pattern recognition, Rejection Option, Support Vector Machines @inproceedings{Luckner2014b, title = {Global and local rejection option in multi-classification task}, author = {Marcin Luckner}, doi = {10.1007/978-3-319-11179-7_61}, issn = {16113349}, year = {2014}, date = {2014-01-01}, booktitle = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)}, volume = {8681 LNCS}, pages = {483--490}, abstract = {This work presents two rejection options. The global rejection option separates the foreign observations - not defined in the classification task - from the normal observations. The local rejection option works after the classification process and separates observations individually for each class. We present implementation of both methods for binary classifiers grouped in a graph structure (tree or directed acyclic graph). Next, we prove that the quality of rejection is identical for both options and depends only on the quality of binary classifiers. The methods are compared on the handwritten digits recognition task. The local rejection option works better for the most part. ? 2014 Springer International Publishing Switzerland.}, keywords = {Graph Ensemble, Pattern recognition, Rejection Option, Support Vector Machines}, pubstate = {published}, tppubtype = {inproceedings} } This work presents two rejection options. The global rejection option separates the foreign observations - not defined in the classification task - from the normal observations. The local rejection option works after the classification process and separates observations individually for each class. We present implementation of both methods for binary classifiers grouped in a graph structure (tree or directed acyclic graph). Next, we prove that the quality of rejection is identical for both options and depends only on the quality of binary classifiers. The methods are compared on the handwritten digits recognition task. The local rejection option works better for the most part. ? 2014 Springer International Publishing Switzerland. |
Luckner, Marcin Global and local rejection option in multi-classification task Inproceedings Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), pp. 483–490, 2014, ISSN: 16113349. Abstract | Links | BibTeX | Tags: Graph Ensemble, Pattern recognition, Rejection Option, Support Vector Machines @inproceedings{Luckner2014e, title = {Global and local rejection option in multi-classification task}, author = {Marcin Luckner}, doi = {10.1007/978-3-319-11179-7_61}, issn = {16113349}, year = {2014}, date = {2014-01-01}, booktitle = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)}, volume = {8681 LNCS}, pages = {483--490}, abstract = {This work presents two rejection options. The global rejection option separates the foreign observations - not defined in the classification task - from the normal observations. The local rejection option works after the classification process and separates observations individually for each class. We present implementation of both methods for binary classifiers grouped in a graph structure (tree or directed acyclic graph). Next, we prove that the quality of rejection is identical for both options and depends only on the quality of binary classifiers. The methods are compared on the handwritten digits recognition task. The local rejection option works better for the most part. ? 2014 Springer International Publishing Switzerland.}, keywords = {Graph Ensemble, Pattern recognition, Rejection Option, Support Vector Machines}, pubstate = {published}, tppubtype = {inproceedings} } This work presents two rejection options. The global rejection option separates the foreign observations - not defined in the classification task - from the normal observations. The local rejection option works after the classification process and separates observations individually for each class. We present implementation of both methods for binary classifiers grouped in a graph structure (tree or directed acyclic graph). Next, we prove that the quality of rejection is identical for both options and depends only on the quality of binary classifiers. The methods are compared on the handwritten digits recognition task. The local rejection option works better for the most part. ? 2014 Springer International Publishing Switzerland. |
2013 |
Luckner, Marcin; Szyszko, Krzysztof RBF ensemble based on reduction of DAG structure Inproceedings Proceedings of the 2013 Federated Conference on Computer Science and Information Systems, pp. 99–105, IEEE, Kraków, 2013. Abstract | Links | BibTeX | Tags: Accuracy, binary classifiers, Chebyshev approximation, classes similarity, Classification, classification cost reduction, DAG structure reduction, Directed Acyclic Graph, directed graphs, Euclidean distance, Glass, Kernel, learning (artificial intelligence), pattern classification, Radial Basis Function, radial basis function ensemble, radial basis function networks, RBF ensemble, recognition accuracy, Support Vector Machines, UCI repository @inproceedings{Luckner2013a, title = {RBF ensemble based on reduction of DAG structure}, author = {Marcin Luckner and Krzysztof Szyszko}, url = {https://fedcsis.org/proceedings/2013/pliks/334.pdf}, year = {2013}, date = {2013-01-01}, booktitle = {Proceedings of the 2013 Federated Conference on Computer Science and Information Systems}, pages = {99--105}, publisher = {IEEE}, address = {Kraków}, abstract = {Binary classifiers are grouped into an ensemble to solve multi-class problems. One of proposed ensemble structure is a directed acyclic graph. In this structure, a classifier is created for each pair of classes. The number of classifiers can be reduced if groups of classes will be separated instead of individual classes. The proposed method is based on the similarity of classes defined as a distance between classes. For near classes the structure of DAG stays immutable. For the distant classes more than one is separated with a single classifier. In this paper, the proposed method is tested in variants based on various metrics. For the tests, several datasets from UCI repository was used and the results were compared with published works. The tests proved that grouping of radial basis functions into such ensemble reduces the classification cost and the recognition accuracy is not reduced significantly.}, keywords = {Accuracy, binary classifiers, Chebyshev approximation, classes similarity, Classification, classification cost reduction, DAG structure reduction, Directed Acyclic Graph, directed graphs, Euclidean distance, Glass, Kernel, learning (artificial intelligence), pattern classification, Radial Basis Function, radial basis function ensemble, radial basis function networks, RBF ensemble, recognition accuracy, Support Vector Machines, UCI repository}, pubstate = {published}, tppubtype = {inproceedings} } Binary classifiers are grouped into an ensemble to solve multi-class problems. One of proposed ensemble structure is a directed acyclic graph. In this structure, a classifier is created for each pair of classes. The number of classifiers can be reduced if groups of classes will be separated instead of individual classes. The proposed method is based on the similarity of classes defined as a distance between classes. For near classes the structure of DAG stays immutable. For the distant classes more than one is separated with a single classifier. In this paper, the proposed method is tested in variants based on various metrics. For the tests, several datasets from UCI repository was used and the results were compared with published works. The tests proved that grouping of radial basis functions into such ensemble reduces the classification cost and the recognition accuracy is not reduced significantly. |
Luckner, Marcin; Szyszko, Krzysztof RBF ensemble based on reduction of DAG structure Inproceedings Proceedings of the 2013 Federated Conference on Computer Science and Information Systems, pp. 99–105, IEEE, Kraków, 2013. Abstract | Links | BibTeX | Tags: Accuracy, binary classifiers, Chebyshev approximation, classes similarity, Classification, classification cost reduction, DAG structure reduction, Directed Acyclic Graph, directed graphs, Euclidean distance, Glass, Kernel, learning (artificial intelligence), pattern classification, Radial Basis Function, radial basis function ensemble, radial basis function networks, RBF ensemble, recognition accuracy, Support Vector Machines, UCI repository @inproceedings{Luckner2013ab, title = {RBF ensemble based on reduction of DAG structure}, author = {Marcin Luckner and Krzysztof Szyszko}, url = {https://fedcsis.org/proceedings/2013/pliks/334.pdf}, year = {2013}, date = {2013-01-01}, booktitle = {Proceedings of the 2013 Federated Conference on Computer Science and Information Systems}, pages = {99--105}, publisher = {IEEE}, address = {Kraków}, abstract = {Binary classifiers are grouped into an ensemble to solve multi-class problems. One of proposed ensemble structure is a directed acyclic graph. In this structure, a classifier is created for each pair of classes. The number of classifiers can be reduced if groups of classes will be separated instead of individual classes. The proposed method is based on the similarity of classes defined as a distance between classes. For near classes the structure of DAG stays immutable. For the distant classes more than one is separated with a single classifier. In this paper, the proposed method is tested in variants based on various metrics. For the tests, several datasets from UCI repository was used and the results were compared with published works. The tests proved that grouping of radial basis functions into such ensemble reduces the classification cost and the recognition accuracy is not reduced significantly.}, keywords = {Accuracy, binary classifiers, Chebyshev approximation, classes similarity, Classification, classification cost reduction, DAG structure reduction, Directed Acyclic Graph, directed graphs, Euclidean distance, Glass, Kernel, learning (artificial intelligence), pattern classification, Radial Basis Function, radial basis function ensemble, radial basis function networks, RBF ensemble, recognition accuracy, Support Vector Machines, UCI repository}, pubstate = {published}, tppubtype = {inproceedings} } Binary classifiers are grouped into an ensemble to solve multi-class problems. One of proposed ensemble structure is a directed acyclic graph. In this structure, a classifier is created for each pair of classes. The number of classifiers can be reduced if groups of classes will be separated instead of individual classes. The proposed method is based on the similarity of classes defined as a distance between classes. For near classes the structure of DAG stays immutable. For the distant classes more than one is separated with a single classifier. In this paper, the proposed method is tested in variants based on various metrics. For the tests, several datasets from UCI repository was used and the results were compared with published works. The tests proved that grouping of radial basis functions into such ensemble reduces the classification cost and the recognition accuracy is not reduced significantly. |
2011 |
Luckner, Marcin Reducing Number of Classifiers in DAGSVM Based on Class Similarity Inproceedings Image Analysis and Processing – ICIAP 2011 Lecture Notes in Computer Science, pp. 514–523, Springer Berlin Heidelberg, 2011. Abstract | Links | BibTeX | Tags: Classification, Directed Acyclic Graph, One–Against–One, Support Vector Machines @inproceedings{Luckner2011a, title = {Reducing Number of Classifiers in DAGSVM Based on Class Similarity}, author = {Marcin Luckner}, url = {http://link.springer.com/chapter/10.1007%2F978-3-642-24085-0_53}, doi = {10.1007/978-3-642-24085-0_53}, year = {2011}, date = {2011-01-01}, booktitle = {Image Analysis and Processing – ICIAP 2011 Lecture Notes in Computer Science}, pages = {514--523}, publisher = {Springer Berlin Heidelberg}, abstract = {Support Vector Machines are excellent binary classifiers. In case of multi–class classification problems individual classifiers can be collected into a directed acyclic graph structure DAGSVM. Such structure implements One-Against-One strategy. In this strategy a split is created for each pair of classes, but, because of hierarchical structure, only a part of them is used in the single classification process. The number of classifiers may be reduced if their classification tasks will be changed from separation of individual classes into separation of groups of classes. The proposed method is based on the similarity of classes. For near classes the structure of DAG stays immutable. For the distant classes more than one is separated with a single classifier. This solution reduces the classification cost. At the same time the recognition accuracy is not reduced in a significant way. Moreover, a number of SV, which influences on the learning time will not grow rapidly.}, keywords = {Classification, Directed Acyclic Graph, One–Against–One, Support Vector Machines}, pubstate = {published}, tppubtype = {inproceedings} } Support Vector Machines are excellent binary classifiers. In case of multi–class classification problems individual classifiers can be collected into a directed acyclic graph structure DAGSVM. Such structure implements One-Against-One strategy. In this strategy a split is created for each pair of classes, but, because of hierarchical structure, only a part of them is used in the single classification process. The number of classifiers may be reduced if their classification tasks will be changed from separation of individual classes into separation of groups of classes. The proposed method is based on the similarity of classes. For near classes the structure of DAG stays immutable. For the distant classes more than one is separated with a single classifier. This solution reduces the classification cost. At the same time the recognition accuracy is not reduced in a significant way. Moreover, a number of SV, which influences on the learning time will not grow rapidly. |
Luckner, Marcin Reducing Number of Classifiers in DAGSVM Based on Class Similarity Inproceedings Image Analysis and Processing – ICIAP 2011 Lecture Notes in Computer Science, pp. 514–523, Springer Berlin Heidelberg, 2011. Abstract | Links | BibTeX | Tags: Classification, Directed Acyclic Graph, One–Against–One, Support Vector Machines @inproceedings{Luckner2011ab, title = {Reducing Number of Classifiers in DAGSVM Based on Class Similarity}, author = {Marcin Luckner}, url = {http://link.springer.com/chapter/10.1007%2F978-3-642-24085-0_53}, doi = {10.1007/978-3-642-24085-0_53}, year = {2011}, date = {2011-01-01}, booktitle = {Image Analysis and Processing – ICIAP 2011 Lecture Notes in Computer Science}, pages = {514--523}, publisher = {Springer Berlin Heidelberg}, abstract = {Support Vector Machines are excellent binary classifiers. In case of multi–class classification problems individual classifiers can be collected into a directed acyclic graph structure DAGSVM. Such structure implements One-Against-One strategy. In this strategy a split is created for each pair of classes, but, because of hierarchical structure, only a part of them is used in the single classification process. The number of classifiers may be reduced if their classification tasks will be changed from separation of individual classes into separation of groups of classes. The proposed method is based on the similarity of classes. For near classes the structure of DAG stays immutable. For the distant classes more than one is separated with a single classifier. This solution reduces the classification cost. At the same time the recognition accuracy is not reduced in a significant way. Moreover, a number of SV, which influences on the learning time will not grow rapidly.}, keywords = {Classification, Directed Acyclic Graph, One–Against–One, Support Vector Machines}, pubstate = {published}, tppubtype = {inproceedings} } Support Vector Machines are excellent binary classifiers. In case of multi–class classification problems individual classifiers can be collected into a directed acyclic graph structure DAGSVM. Such structure implements One-Against-One strategy. In this strategy a split is created for each pair of classes, but, because of hierarchical structure, only a part of them is used in the single classification process. The number of classifiers may be reduced if their classification tasks will be changed from separation of individual classes into separation of groups of classes. The proposed method is based on the similarity of classes. For near classes the structure of DAG stays immutable. For the distant classes more than one is separated with a single classifier. This solution reduces the classification cost. At the same time the recognition accuracy is not reduced in a significant way. Moreover, a number of SV, which influences on the learning time will not grow rapidly. |
2008 |
Luckner, Marcin Comparison of hierarchical svm structures in letters recognition task Inproceedings IEEE CIS-Poland Chapter Edited Volume, Warsaw, Poland, 2008. Links | BibTeX | Tags: decisions tree, hierar-, Pattern recognition, Support Vector Machines @inproceedings{Luckner2008a, title = {Comparison of hierarchical svm structures in letters recognition task}, author = {Marcin Luckner}, url = {http://www.academia.edu/8882762/Comparison_of_hierarchical_svm_structures_in_letters_recognition_task}, year = {2008}, date = {2008-01-01}, booktitle = {IEEE CIS-Poland Chapter Edited Volume}, address = {Warsaw, Poland}, keywords = {decisions tree, hierar-, Pattern recognition, Support Vector Machines}, pubstate = {published}, tppubtype = {inproceedings} } |
Luckner, Marcin Comparison of hierarchical svm structures in letters recognition task Inproceedings IEEE CIS-Poland Chapter Edited Volume, Warsaw, Poland, 2008. Links | BibTeX | Tags: decisions tree, hierar-, Pattern recognition, Support Vector Machines @inproceedings{Luckner2008ab, title = {Comparison of hierarchical svm structures in letters recognition task}, author = {Marcin Luckner}, url = {http://www.academia.edu/8882762/Comparison_of_hierarchical_svm_structures_in_letters_recognition_task}, year = {2008}, date = {2008-01-01}, booktitle = {IEEE CIS-Poland Chapter Edited Volume}, address = {Warsaw, Poland}, keywords = {decisions tree, hierar-, Pattern recognition, Support Vector Machines}, pubstate = {published}, tppubtype = {inproceedings} } |
Publications
2014 |
Pattern recognition with rejection: Application to handwritten digits Inproceedings 2014 4th World Congress on Information and Communication Technologies (WICT 2014), pp. 326–331, IEEE, 2014, ISBN: 978-1-4799-8115-1. |
Pattern recognition with rejection: Application to handwritten digits Inproceedings 2014 4th World Congress on Information and Communication Technologies (WICT 2014), pp. 326–331, IEEE, 2014, ISBN: 978-1-4799-8115-1. |
Global and local rejection option in multi-classification task Inproceedings Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), pp. 483–490, 2014, ISSN: 16113349. |
Global and local rejection option in multi-classification task Inproceedings Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), pp. 483–490, 2014, ISSN: 16113349. |
2013 |
RBF ensemble based on reduction of DAG structure Inproceedings Proceedings of the 2013 Federated Conference on Computer Science and Information Systems, pp. 99–105, IEEE, Kraków, 2013. |
RBF ensemble based on reduction of DAG structure Inproceedings Proceedings of the 2013 Federated Conference on Computer Science and Information Systems, pp. 99–105, IEEE, Kraków, 2013. |
2011 |
Reducing Number of Classifiers in DAGSVM Based on Class Similarity Inproceedings Image Analysis and Processing – ICIAP 2011 Lecture Notes in Computer Science, pp. 514–523, Springer Berlin Heidelberg, 2011. |
Reducing Number of Classifiers in DAGSVM Based on Class Similarity Inproceedings Image Analysis and Processing – ICIAP 2011 Lecture Notes in Computer Science, pp. 514–523, Springer Berlin Heidelberg, 2011. |
2008 |
Comparison of hierarchical svm structures in letters recognition task Inproceedings IEEE CIS-Poland Chapter Edited Volume, Warsaw, Poland, 2008. |
Comparison of hierarchical svm structures in letters recognition task Inproceedings IEEE CIS-Poland Chapter Edited Volume, Warsaw, Poland, 2008. |