SV REDUCTION METHOD FOR MULTI-CLASS SVM

摘要

PPROBLEM TO BE SOLVED: To provide a support vector (SV) reduction method for a multi-class support vector machine (SVM) reducing the number of SVs contained in a multi-class SVM without becoming trapped in a local optimum solution at a high speed and with a high degree of precision. PSOLUTION: The SV reduction method for a multi-class SVM includes: a step S20 of selecting, from a plurality of initially present support vectors, support vector pairs zSBi/SBand zSBj/SB(i, j=1, 2, ..., NSBs/SB); a step of preparing a single-variable objective function with a single maximum to determine a maximum value k of the objective function; and a step S30 for applying the maximum value k to the support vector pairs zSBi/SBand zSBj/SBto determine a temporary vector Ztemp [i] of small classification errors. The support vector pairs zSBi/SBand zSBj/SBare represented by the temporary vector Ztemp [i]. PCOPYRIGHT: (C)2010,JPO&INPIT