学科分类
/ 1
1 个结果
  • 简介:Toacceleratethetrainingofsupportvectordomaindescription(SVDD),confidencesupportvectordomaindescription(CSVDD)isproposedbasedontheobservationthatthedescriptionboundaryisdeterminedbyasmallsubsetoftrainingdatacalledsupportvectors.Namely,thenumberoftrainingsamplesintheuserdefinedsphereiscalculatedandtakenastheconfidencemeasure,accordingtowhichthetrainingsamplesarerankedinascendingorder.ThoseformerrankedonesareselectedastheboundarytargetsfortheSVDDtraining.SimulationsonUCIdatademonstratetheeffectivenessandsuperiorityofCSVDD:thenumberoftrainingtargetsandthetrainingtimearereducedwithoutanylossofaccuracy.

  • 标签: 无线电导引系统 CSVDD SVDD 系统工程