Font Size: a A A

Support Vector Machine Model Research And Application

Posted on:2010-10-03Degree:MasterType:Thesis
Country:ChinaCandidate:H X LiangFull Text:PDF
GTID:2178360302962635Subject:Computer applications
Abstract/Summary:PDF Full Text Request
SVM is a new machine learning technology which was developed in middle of 90th. Different from traditional neural network technology, SVM is based on statistics learning theory (SLT).Sufficient sample is a necessary prior condition for traditional statistics learning, but SLT focus on small sample and its regulation or study method, which institutes a robust theoretical frame to machine learning. Practice shows that due to simple structure and high technical performance especially to generalization ability, SVM solves small sample problem more powerful in reality. At present, SVM becomes a hot topic among machine learning area in internationality.This paper mainly discusses several aspects as follows:1)By studying and analyzing fuzzy set theory and probability theory, the paper constitution a model named probability fuzzy support vector machine(PFSVM),which can reduce or vanish the affection of outlier or noisy data to the whole training sample, also makes up for deficiencies of FSVM. New model PFSVM considers clustering characteristic as well as probability distribution of the data, fully reflects the different roles of data samples to obtain a more reasonable classification hyperplane.2)Applying PFSVM to image retrieval could solve small sample problem in content-based image retrieval relevance feedback process. The new advance PL-PFSVM algorithm helps to solve not only small sample problem in relevance feedback but also effectively apply to complicated semantic image retrieval. Membership and probability value controlling PFSVM to train and study together makes a high performance in image retrieval feedback.3)To obtain cryptic information of sample, this paper proposes a new model--- rough support vector machine (RSVM) which well combines rough set theory and support vector machine theory. It can clearly embody equivalent information of data though adopting the RSVM.
Keywords/Search Tags:support vector machine, probability, fuzzy, rough set, relevance feedback
PDF Full Text Request
Related items