半监督多标记学习的基因功能分析_陈晓峰.pdfVIP

3 1 V ol . 3 l . 1 200 8 2 CA A I T ransact io ns on Int ellig ent Sy st ems F eb . 200 8 1 1 1, 2 陈晓峰 , 王士同, 曹苏群 ( 1. 江南大学 信息工程学院, 江苏 无锡 2 14 122; 2. 淮阴工学院 机械系, 江苏 淮安 223001 : , . , , , , . . , , , . SM L _ SV M . , SM L _ SV M PT 4 , , ( M A P K , , SVM . y east genbase , SM L _SV M PT 4 M LSV M M L SVM . : ; ; ; : T P 18 1 : A : 167 3- 4785( 200 8 01- 00 83- 08 Gene function analysis of sem-i supervised mult-i label learning 1 1 1, 2 CH EN Xiao-feng , W A N G Sh-i t ong , CA O Su- qu n ( 1. Scho ol of Infor mation T echno lo gy , Jiang nan U niv ersity , Wuxi 214122 , China; 2. Department o f M echanical Engineer ing , Huaiy in Institut e o f T echno log y, H uai. an 22300 1, China Abstract:Co nv ent io nal machine learning is u sed only f or single label lear ning, implying t hat ev ery sample has o nly one label. H ow ever, in bioinfo rmat ics, a gene has more t han o ne f unct ion , so it needs mo re than one label. T heref ore, mult-i label learning is mor e ef fect ive fo r ident ify ing gene gr oup s t han convent ional learning appro ach . Current r esearch mainly fo cu ses on supervised mult-i label learning . T he problem of ef- f ect ive sem-i superv ised multi- label learning st rat egies f or labeled exam ples and unlabeled ex amples of g ene ex pression dat asets st ill remains unsolved. In t his paper, a sem-i supervised m ult-i label learning algor it hm, named SML_SV M , is presented as an ef