Concept acquisition is one of the main tasks in semantic-based image retrieval. Traditional concept acquisition methods generally involve two aspects: unsupervised clustering and supervised classification. In the classification, SVM was widely concerned for its excellent performance. However, there are many unlabeled samples in application environment. If they are fully utilized, a more precise classification will be obtained. In this paper, a novel semi-supervised fuzzy SVM clustering framework was presented. In the framework, the spatial distribution information of the unlabeled samples and the prompted information of the labeled samples are integrated to obtain better results. The contributions of this paper are as follows: (1) proposed a semi-supervised fuzzy SVM clustering framework; (2) discussed the acquisition of hyper-planes for data with different importance.
[1]
Farid Melgani,et al.
A Multiobjective Genetic SVM Approach for Classification Problems With Limited Training Samples
,
2009,
IEEE Transactions on Geoscience and Remote Sensing.
[2]
Daoqiang Zhang,et al.
A simultaneous learning framework for clustering and classification
,
2009,
Pattern Recognit..
[3]
Cheng-Jia Li.
Robust Extension of FCM Algorithm
,
2006,
2006 International Conference on Machine Learning and Cybernetics.
[4]
Zilu Ying,et al.
Fusing binary support vector machines (SVM) into multiclass SVM
,
2006,
International Symposium on Instrumentation and Control Technology.