一種基于深度玻爾茲曼機(jī)的半監(jiān)督典型相關(guān)分析算法
發(fā)布時(shí)間:2019-03-14 18:55
【摘要】:從模式分類(lèi)的角度出發(fā),針對(duì)典型相關(guān)分析(canonical correlation analysis,CCA)算法不適應(yīng)于高層次關(guān)聯(lián)的缺陷,提出了改進(jìn)算法。將深度學(xué)習(xí)理論與典型相關(guān)分析算法相結(jié)合,基于深度玻爾茲曼機(jī)理論提出了一種半監(jiān)督典型相關(guān)分析算法。通過(guò)深度玻爾茲曼機(jī)提取出樣本的顯層特征與隱層特征,結(jié)合已標(biāo)注樣本的監(jiān)督信息,構(gòu)造出最有效的鑒別特征。依據(jù)ORL、Yale和AR人臉數(shù)據(jù)庫(kù)進(jìn)行仿真實(shí)驗(yàn),實(shí)驗(yàn)結(jié)果表明:本文算法與其他的方法相比,具有更好的識(shí)別效果。
[Abstract]:From the point of view of pattern classification, an improved (canonical correlation analysis,CCA (canonical correlation analysis) algorithm is proposed to solve the problem that it is not suitable for high-level association. Based on the deep Boltzmann machine theory, a semi-supervised canonical correlation analysis algorithm is proposed based on the combination of depth learning theory and typical correlation analysis algorithm. The most effective discriminant features are constructed by using depth Boltzmann machine to extract the explicit and hidden layer features of the samples and combine the supervised information of the labeled samples to construct the most effective discriminant features. The simulation results based on ORL,Yale and AR face database show that the proposed algorithm has better recognition performance than other methods.
【作者單位】: 鄭州大學(xué)信息工程學(xué)院;
【基金】:國(guó)家自然科學(xué)基金項(xiàng)目(61210005;61331021)
【分類(lèi)號(hào)】:TP391.41
本文編號(hào):2440268
[Abstract]:From the point of view of pattern classification, an improved (canonical correlation analysis,CCA (canonical correlation analysis) algorithm is proposed to solve the problem that it is not suitable for high-level association. Based on the deep Boltzmann machine theory, a semi-supervised canonical correlation analysis algorithm is proposed based on the combination of depth learning theory and typical correlation analysis algorithm. The most effective discriminant features are constructed by using depth Boltzmann machine to extract the explicit and hidden layer features of the samples and combine the supervised information of the labeled samples to construct the most effective discriminant features. The simulation results based on ORL,Yale and AR face database show that the proposed algorithm has better recognition performance than other methods.
【作者單位】: 鄭州大學(xué)信息工程學(xué)院;
【基金】:國(guó)家自然科學(xué)基金項(xiàng)目(61210005;61331021)
【分類(lèi)號(hào)】:TP391.41
【相似文獻(xiàn)】
相關(guān)期刊論文 前3條
1 劉來(lái)福,唐志宇,,匡錦瑜;向量玻爾茲曼機(jī)[J];北京師范大學(xué)學(xué)報(bào)(自然科學(xué)版);1996年04期
2 秦勝君;盧志平;;基于限制玻爾茲曼機(jī)的無(wú)極性標(biāo)注情感分類(lèi)研究[J];科學(xué)技術(shù)與工程;2013年35期
3 ;[J];;年期
相關(guān)碩士學(xué)位論文 前5條
1 李娟;基于RBM的小分子活性及選擇性研究[D];蘭州大學(xué);2015年
2 葉睿;基于深度學(xué)習(xí)的人臉檢測(cè)方法研究[D];哈爾濱工業(yè)大學(xué);2015年
3 王海麟;通過(guò)信息幾何方法挖掘玻爾茲曼機(jī)的不變性[D];天津大學(xué);2014年
4 仝少敏;基于受限玻爾茲曼機(jī)的面部運(yùn)動(dòng)識(shí)別方法研究[D];哈爾濱工業(yè)大學(xué);2014年
5 李平;監(jiān)督概率主題模型研究[D];安徽工業(yè)大學(xué);2014年
本文編號(hào):2440268
本文鏈接:http://www.sikaile.net/kejilunwen/ruanjiangongchenglunwen/2440268.html
最近更新
教材專(zhuān)著