CAJ | 학술논문

针对目前支持向量机（SVM）中混合核函数的不足，提出一种自适应加权混合核函数。该核函数能自适应调节新映射空间样本点的距离，改变序列最小优化（SMO）过程中的修正因子，以削弱惩罚因子的影响，改变拉格朗日乘子的取值，优化支持向量的选取，进而获得更优的分类界面，提高SVM的分类能力，并首次提出将混合核函数SVM应用于脑肿瘤分割。实验结果表明，该方法能更高效准确地分割脑肿瘤。
침대목전지지향량궤（SVM）중혼합핵함수적불족，제출일충자괄응가권혼합핵함수。해핵함수능자괄응조절신영사공간양본점적거리，개변서렬최소우화（SMO）과정중적수정인자，이삭약징벌인자적영향，개변랍격랑일승자적취치，우화지지향량적선취，진이획득경우적분류계면，제고SVM적분류능력，병수차제출장혼합핵함수SVM응용우뇌종류분할。실험결과표명，해방법능경고효준학지분할뇌종류。
Aiming at the deficiency of combined-kernel function in current support vector machine (SVM),we present an adaptive weighted combined-kernel function.This kernel function is able to adaptively adjust the distance of sample points in new mapping space, changes the correction factor in sequential minimal optimisation (SMO)process to weaken the influence of penalty factor,and changes the value of Lagrange multiplier and optimises the selection of support vectors as well,so as to get a better classification interaction and to improve the classification ability of SVM.Furthermore,we propose the first time to apply the combined-kernel function SVM in brain tumour segmentation.Experimental results show that the method can more effectively segment the brain tumour.