一種選擇特征的稀疏在線學(xué)習(xí)算法
發(fā)布時間:2019-07-10 16:54
【摘要】:為了有效處理海量、高維、稀疏的大數(shù)據(jù),提高對數(shù)據(jù)的分類效率,提出一種基于L_1準則稀疏性原理的在線學(xué)習(xí)算法(a sparse online learning algorithm for selection feature,SFSOL)。運用在線機器學(xué)習(xí)算法框架,對高維流式數(shù)據(jù)的特征進行新穎的"取整"處理,加大數(shù)據(jù)特征稀疏性的同時增強了閥值范圍內(nèi)部分特征的值,極大地提高了對稀疏數(shù)據(jù)分類的效果。利用公開的數(shù)據(jù)集對SFSOL算法的性能進行分析,并將該算法與其它3種稀疏在線學(xué)習(xí)算法的性能進行比較,試驗結(jié)果表明提出的SFSOL算法對高維稀疏數(shù)據(jù)分類的準確性更高。
[Abstract]:In order to deal with massive, high-dimensional and sparse big data effectively and improve the classification efficiency of data, an online learning algorithm (a sparse online learning algorithm for selection feature,SFSOL based on the sparse principle of L 鈮,
本文編號:2512755
[Abstract]:In order to deal with massive, high-dimensional and sparse big data effectively and improve the classification efficiency of data, an online learning algorithm (a sparse online learning algorithm for selection feature,SFSOL based on the sparse principle of L 鈮,
本文編號:2512755
本文鏈接:http://sikaile.net/kejilunwen/zidonghuakongzhilunwen/2512755.html
最近更新
教材專著