现代计算机(专业版)
現代計算機(專業版)
현대계산궤(전업판)
MODERN COMPUTER
2014年
6期
25-29,34
,共6页
支持向量机%参数优化%遗传算法%分裂选择
支持嚮量機%參數優化%遺傳算法%分裂選擇
지지향량궤%삼수우화%유전산법%분렬선택
SVM%Parameter Optimization%Genetic Algorithm%Disruptive Selection
支持向量机中惩罚参数和核参数的取值对其性能影响较大,但在最优参数选取上一直缺少理论指导。依据分裂选择的思想对遗传算法的选择操作做改进,以保持种群的多样性,同时利用支持向量计数法构造适应度函数,用改进后的遗传算法对支持向量机进行参数优化,不仅在一定程度上避免过早陷入局部最优,还能提高计算效率。实验表明,用改进的遗传算法优化支持向量机的参数耗时更短,并且得到的支持向量机有更好的分类效果。
支持嚮量機中懲罰參數和覈參數的取值對其性能影響較大,但在最優參數選取上一直缺少理論指導。依據分裂選擇的思想對遺傳算法的選擇操作做改進,以保持種群的多樣性,同時利用支持嚮量計數法構造適應度函數,用改進後的遺傳算法對支持嚮量機進行參數優化,不僅在一定程度上避免過早陷入跼部最優,還能提高計算效率。實驗錶明,用改進的遺傳算法優化支持嚮量機的參數耗時更短,併且得到的支持嚮量機有更好的分類效果。
지지향량궤중징벌삼수화핵삼수적취치대기성능영향교대,단재최우삼수선취상일직결소이론지도。의거분렬선택적사상대유전산법적선택조작주개진,이보지충군적다양성,동시이용지지향량계수법구조괄응도함수,용개진후적유전산법대지지향량궤진행삼수우화,불부재일정정도상피면과조함입국부최우,환능제고계산효솔。실험표명,용개진적유전산법우화지지향량궤적삼수모시경단,병차득도적지지향량궤유경호적분류효과。
The selection of support vector machine's punished parameter and kernel parameter will affect its performance greatly and there are no theory for guiding the selection of its optimal parameters. Improves selection operation of genetic algorithm to maintain the diversity of the population based on the idea of disruptive selection. Applies the improved genetic algorithm for optimizing SVM parameters, and uses the support vectors counting method to construct the fitness function, which avoid falling into local optimum in some extent and reduce the computation greatly. Experiments show that the improved algorithm can optimize the parameters of SVM with shorten running time, and the support vector machine will get better classification performance.