湖南大学学报(自然科学版)
湖南大學學報(自然科學版)
호남대학학보(자연과학판)
JOURNAL OF HUNAN UNIVERSITY(NATURAL SCIENCES EDITION)
2010年
4期
37-41
,共5页
最小二乘支持向量机%学习算法%稀疏性%选择性删除%系统建模
最小二乘支持嚮量機%學習算法%稀疏性%選擇性刪除%繫統建模
최소이승지지향량궤%학습산법%희소성%선택성산제%계통건모
least squares support vector machine(LSSVM)%learning algorithms%sparsity%selective deletion%system modelling
为了减少在线最小二乘支持向量机(LSSVM)的计算量和存储空间,提出了一种在线稀疏LSSVM.这种LSSVM利用滑动时间窗中部分时刻的样本作为训练样本集.新时刻的样本总是加入训练样本集;每次删除样本时,若滑动时间窗最前端时刻的样本在训练样本集中,则删除它,否则从训练样本集中选择留一法预测误差最小的样本删除.与现有的在线LSSVM相比,这种在线稀疏LSSVM能用较少的样本学习系统较多的特性,能提高时空效率;与现有的在线稀疏LSSVM相比,它能摆脱陈旧样本的影响,更加适应系统的时变性.系统建模仿真实验表明,该在线稀疏LSSVM能节省时间和空间,具有较高的预测精度.
為瞭減少在線最小二乘支持嚮量機(LSSVM)的計算量和存儲空間,提齣瞭一種在線稀疏LSSVM.這種LSSVM利用滑動時間窗中部分時刻的樣本作為訓練樣本集.新時刻的樣本總是加入訓練樣本集;每次刪除樣本時,若滑動時間窗最前耑時刻的樣本在訓練樣本集中,則刪除它,否則從訓練樣本集中選擇留一法預測誤差最小的樣本刪除.與現有的在線LSSVM相比,這種在線稀疏LSSVM能用較少的樣本學習繫統較多的特性,能提高時空效率;與現有的在線稀疏LSSVM相比,它能襬脫陳舊樣本的影響,更加適應繫統的時變性.繫統建模倣真實驗錶明,該在線稀疏LSSVM能節省時間和空間,具有較高的預測精度.
위료감소재선최소이승지지향량궤(LSSVM)적계산량화존저공간,제출료일충재선희소LSSVM.저충LSSVM이용활동시간창중부분시각적양본작위훈련양본집.신시각적양본총시가입훈련양본집;매차산제양본시,약활동시간창최전단시각적양본재훈련양본집중,칙산제타,부칙종훈련양본집중선택류일법예측오차최소적양본산제.여현유적재선LSSVM상비,저충재선희소LSSVM능용교소적양본학습계통교다적특성,능제고시공효솔;여현유적재선희소LSSVM상비,타능파탈진구양본적영향,경가괄응계통적시변성.계통건모방진실험표명,해재선희소LSSVM능절성시간화공간,구유교고적예측정도.
To reduce the computation time and the storage space of online least squares support vector machine (LSSVM), an online sparse LSSVM was proposed. This LSSVM only takes samples at partial moments among sliding time window as training samples set (TSS). The new sample is learned necessarily. When sample elimination is performed, if the sample at the oldest moment among sliding time window exists in TSS, it will be removed during decremental learning. Otherwise, the sample with the smallest leave-one-out predicting error among TSS is selected and deleted. Compared with the existing online LSSVM, the proposed online sparse LSSVM can learn more characteristic of the system with fewer samples, and heighten time-space efficiency. Compared with the existing online spare LSSVM, it can get rid of the obsolete sample, and better adapt to time-variant properties of system. Numerical simulation results for system modeling have shown that the proposed online sparse LSSVM can save time and space, and provide accurate predictions.