A matrix modular SVM robust to imbalanced data for efficient visual concept detection

Authors:
Herve Glotin;Zhong-Qiu Zhao;Jun Gao;Xindong Wu
Affiliations:
Sciences and Information Lab. (LSIS), UMR CNRS, La Garde, France;College of Computer Science and Information Engineering, Hefei University of Technology, China, Hefei, China;College of Computer Science and Information Engineering, Hefei University of Technology, China, Hefei, China;Department of Computer Science, University of Vermont, USA, Vermont, USA
Venue:
Proceedings of the international conference on Multimedia information retrieval
Year:
2010

Citing 7
Cited 1

The nature of statistical learning theory

The nature of statistical learning theory
Least Squares Support Vector Machine Classifiers

Neural Processing Letters
Generalized Discriminant Analysis Using a Kernel Approach

Neural Computation
Letters: Palmprint recognition with 2DPCA+PCA based on modular neural networks

Neurocomputing
Solving multiclass learning problems via error-correcting output codes

Journal of Artificial Intelligence Research
Efficient image concept indexing by harmonic & arithmetic profiles entropy

ICIP'09 Proceedings of the 16th IEEE international conference on Image processing
Efficient classification for multiclass problems using modular neural networks

IEEE Transactions on Neural Networks

Rule-Based Semantic Concept Classification from Large-Scale Video Collections

International Journal of Multimedia Data Engineering & Management

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, we propose a novel matrix modular support vector Machine (MMSVM) classifier that partitions an image retrieval task into many easier two-class tasks between subsets, each of which is accomplished by a SVM model, and then combines the outputs of the SVM models to produce the final decision. The classifier is tested on ImageClef2009 Photo Annotation, with a comparison with the single SVM model. The experimental results show that our MMSVM model performs well as a classifier in image retrieval, especially in enhancing the classification accuracy for positive samples. We also demonstrate that the MMSVM model has an apparent complementary classification capability to SVM. A good fusion on them might improve the accuracy of image retrieval.