TY - JOUR
T1 - A genetic-based subspace analysis method for improving Error-Correcting Output Coding
AU - Bagheri, Mohammad Ali
AU - Gao, Qigang
AU - Escalera, Sergio
N1 - Funding Information:
The authors would like to thank Ms. Fatemeh Yazdanpanah for her kind support in conducting experiments. Mohammad Ali Bagheri Bagheri was born in Shiraz, Iran, in 1983. He received his B.S. degree in Industrial Engineering from Sharif University of Technology in 2005 and his M.S. degree in Information Technology from Tarbiat Modares University, Tehran, Iran, in 2008. He is currently a research assistant at Faculty of Computer Science, Dalhousie University, Canada. His research interests include pattern recognition, machine learning with special interest in machine olfaction, image analysis, and evolutionary computation. Qigang Gao is currently a professor of Computer Science at Dalhousie University, Canada (1993–). Gao got his undergraduate education in Xi'an Jiao Tong University, China (1973–1976), and received both M.A.Sc. and Ph.D. degrees from the University of Waterloo, Canada (1988 and 1993, respectively). His research interests include perceptual knowledge based visual pattern recognition, data mining and data warehouses, and web-based intelligent information systems. He has supervised/co-supervised 44 graduate students at both Ph.D. and Masters levels, and contributed to 75 research articles (authored/co-authored) in the related areas. Sergio Escalera received the B.S. and M.S. degrees from the Universitat Autònoma de Barcelona (UAB), Barcelona, Spain, in 2003 and 2005, respectively. He obtained the Ph.D. degree on Multi-class visual categorization systems at Computer Vision Center, UAB. He obtained the 2008 best Thesis award on informatics at Universitat Autònoma de Barcelona. He is a lecturer of the Department of Applied Mathematics and Analysis, Universitat de Barcelona. He is a partial time professor at Universitat Oberta de Catalunya. He is the member of the Computer Vision Center (UAB). He is a member of the Perceptual Computing Group and a consolidated research group of Catalonia. He is also a member of the International Foundation of Research & Analysis. He is an editorial board member of Journal of Convergence Section C: Web and Multimedia, WebmedCentral, and American Journal of Intelligent Systems. His research interests include, between others, machine learning, statistical pattern recognition, visual object recognition, and human computer interaction systems, with special interest in human pose recovery and behavior analysis.
PY - 2013/10
Y1 - 2013/10
N2 - Two key factors affecting the performance of Error Correcting Output Codes (ECOC) in multiclass classification problems are the independence of binary classifiers and the problem-dependent coding design. In this paper, we propose an evolutionary algorithm-based approach to the design of an application-dependent codematrix in the ECOC framework. The central idea of this work is to design a three-dimensional codematrix, where the third dimension is the feature space of the problem domain. In order to do that, we consider the feature space in the design process of the codematrix with the aim of improving the independence and accuracy of binary classifiers. The proposed method takes advantage of some basic concepts of ensemble classification, such as diversity of classifiers, and also benefits from the evolutionary approach for optimizing the three-dimensional codematrix, taking into account the problem domain. We provide a set of experimental results using a set of benchmark datasets from the UCI Machine Learning Repository, as well as two real multiclass Computer Vision problems. Both sets of experiments are conducted using two different base learners: Neural Networks and Decision Trees. The results show that the proposed method increases the classification accuracy in comparison with the state-of-the-art ECOC coding techniques.
AB - Two key factors affecting the performance of Error Correcting Output Codes (ECOC) in multiclass classification problems are the independence of binary classifiers and the problem-dependent coding design. In this paper, we propose an evolutionary algorithm-based approach to the design of an application-dependent codematrix in the ECOC framework. The central idea of this work is to design a three-dimensional codematrix, where the third dimension is the feature space of the problem domain. In order to do that, we consider the feature space in the design process of the codematrix with the aim of improving the independence and accuracy of binary classifiers. The proposed method takes advantage of some basic concepts of ensemble classification, such as diversity of classifiers, and also benefits from the evolutionary approach for optimizing the three-dimensional codematrix, taking into account the problem domain. We provide a set of experimental results using a set of benchmark datasets from the UCI Machine Learning Repository, as well as two real multiclass Computer Vision problems. Both sets of experiments are conducted using two different base learners: Neural Networks and Decision Trees. The results show that the proposed method increases the classification accuracy in comparison with the state-of-the-art ECOC coding techniques.
KW - Ensemble classification
KW - Error Correcting Output Codes
KW - Evolutionary computation
KW - Feature subspace
KW - Multiclass classification
UR - http://www.scopus.com/inward/record.url?scp=84878011646&partnerID=8YFLogxK
U2 - 10.1016/j.patcog.2013.03.014
DO - 10.1016/j.patcog.2013.03.014
M3 - Journal article
AN - SCOPUS:84878011646
SN - 0031-3203
VL - 46
SP - 2830
EP - 2839
JO - Pattern Recognition
JF - Pattern Recognition
IS - 10
ER -