【摘要】：目前的遷移學(xué)習(xí)模型旨在利用事先準(zhǔn)備好的源域數(shù)據(jù)為目標(biāo)域?qū)W習(xí)提供輔助知識(shí),即從源域抽象出與目標(biāo)域共享的知識(shí)結(jié)構(gòu)時(shí),使用所有的源域數(shù)據(jù)。然而,由于人力資源的限制,收集真實(shí)場(chǎng)景下整體與目標(biāo)域相關(guān)的源域數(shù)據(jù)并不現(xiàn)實(shí)。提出了一種泛化的經(jīng)驗(yàn)風(fēng)險(xiǎn)最小化選擇性知識(shí)利用模型,并給出了該模型的理論風(fēng)險(xiǎn)上界。所提模型能夠自動(dòng)篩選出與目標(biāo)域相關(guān)的源域數(shù)據(jù)子集,解決了源域只有部分知識(shí)可用的問(wèn)題,進(jìn)而避免了在真實(shí)場(chǎng)景下使用整個(gè)源域數(shù)據(jù)集帶來(lái)的負(fù)遷移效應(yīng)。在模擬數(shù)據(jù)集和真實(shí)數(shù)據(jù)集上進(jìn)行了仿真實(shí)驗(yàn),結(jié)果顯示所提算法較之傳統(tǒng)遷移學(xué)習(xí)算法性能更佳。域相關(guān)的源域數(shù)據(jù)并不現(xiàn)實(shí)。提出了一種泛化的經(jīng)驗(yàn)風(fēng)險(xiǎn)最小化選擇性知識(shí)利用模型,并給出了該模型的理論風(fēng)險(xiǎn)上界。所提模型能夠自動(dòng)篩選出與目標(biāo)域相關(guān)的源域數(shù)據(jù)子集,解決了源域只有部分知識(shí)可用的問(wèn)題,進(jìn)而避免了在真實(shí)場(chǎng)景下使用整個(gè)源域數(shù)據(jù)集帶來(lái)的負(fù)遷移效應(yīng)。在模擬數(shù)據(jù)集和真實(shí)數(shù)據(jù)集上進(jìn)行了仿真實(shí)驗(yàn),結(jié)果顯示所提算法較之傳統(tǒng)遷移學(xué)習(xí)算法性能更佳。
[Abstract]:The current transfer learning model aims to provide auxiliary knowledge for target domain learning using pre-prepared source domain data, that is, when abstracting the knowledge structure shared with the target domain from the source domain, all the source domain data are used. However, due to the limitation of human resources, it is not realistic to collect the source domain data related to the target domain in the real scene. In this paper, a generalized empirical risk minimization selective knowledge utilization model is proposed, and the upper bound of the theoretical risk of the model is given. The proposed model can automatically filter out the subset of source domain data related to the target domain, which solves the problem that only part of the knowledge is available in the source domain, and thus avoids the negative migration effect brought by using the whole source domain data set in the real scene. Simulation experiments on simulated data sets and real data sets show that the proposed algorithm performs better than the traditional migration learning algorithm. Domain-related source domain data is not realistic. In this paper, a generalized empirical risk minimization selective knowledge utilization model is proposed, and the upper bound of the theoretical risk of the model is given. The proposed model can automatically filter out the subset of source domain data related to the target domain, which solves the problem that only part of the knowledge is available in the source domain, and thus avoids the negative migration effect brought by using the whole source domain data set in the real scene. Simulation experiments on simulated data sets and real data sets show that the proposed algorithm performs better than the traditional migration learning algorithm.
【作者單位】： 江南大學(xué)數(shù)字媒體學(xué)院;
【基金】：國(guó)家自然科學(xué)基金No.61272210~~
【分類號(hào)】：TP181

【參考文獻(xiàn)】

相關(guān)期刊論文 前3條

1 杭文龍;蔣亦樟;劉解放;王士同;;遷移近鄰傳播聚類算法[J];軟件學(xué)報(bào);2016年11期

2 毛發(fā)貴;李碧雯;沈備軍;;基于實(shí)例遷移的跨項(xiàng)目軟件缺陷預(yù)測(cè)[J];計(jì)算機(jī)科學(xué)與探索;2016年01期

3 楊昌健;鄧趙紅;蔣亦樟;王士同;;基于遷移學(xué)習(xí)的癲癇EEG信號(hào)自適應(yīng)識(shí)別[J];計(jì)算機(jī)科學(xué)與探索;2014年03期

【共引文獻(xiàn)】

相關(guān)期刊論文 前6條

1 龔凱樂(lè);成穎;孫建軍;;基于參與者共現(xiàn)分析的博文聚類研究[J];現(xiàn)代圖書(shū)情報(bào)技術(shù);2016年10期

2 謝麗瀟;鄧趙紅;史熒中;王士同;;面向癲癇EEG自適應(yīng)識(shí)別的遷移徑向基神經(jīng)網(wǎng)絡(luò)[J];計(jì)算機(jī)科學(xué)與探索;2016年12期

3 程e，

本文編號(hào)：2255002