劉婧媛，廈門大學(xué)經(jīng)濟(jì)學(xué)院統(tǒng)計(jì)學(xué)與數(shù)據(jù)科學(xué)系教授、博士生導(dǎo)師，入選國(guó)家級(jí)高層次人才，廈門大學(xué)南強(qiáng)卓越教學(xué)名師，廈門大學(xué)南強(qiáng)青年拔尖人才A類。美國(guó)賓夕法尼亞州立大學(xué)統(tǒng)計(jì)學(xué)博士畢業(yè)。科研方面主要從事高維及復(fù)雜數(shù)據(jù)的統(tǒng)計(jì)方法、網(wǎng)絡(luò)數(shù)據(jù)建模與推斷、統(tǒng)計(jì)基因?qū)W等領(lǐng)域的工作，在JASA，JOE, JBES等國(guó)際權(quán)威學(xué)術(shù)期刊發(fā)表論文30余篇，擔(dān)任AOAS等權(quán)威期刊編委，入選福建省杰出青年科研人才計(jì)劃。教學(xué)方面曾獲國(guó)家級(jí)一流課程、國(guó)家級(jí)教學(xué)成果二等獎(jiǎng)（團(tuán)體）、福建省教學(xué)成果特等獎(jiǎng)、福建省創(chuàng)新教學(xué)比賽二等獎(jiǎng)、廈門大學(xué)“我最喜愛的十位教師”、廈門大學(xué)教學(xué)比賽特等獎(jiǎng)等榮譽(yù)。
報(bào)告摘要：The main challenge that sets transfer learning apart from traditional supervised learning is the distribution shift, reflected as the shift between the source and target models and that between the marginal covariate distributions. High-dimensional data introduces unique challenges, such as covariate shifts in the covariate correlation structure and model shifts across individual features in the model. In this work, we tackle model shifts in the presence of covariate shifts in the high-dimensional regression setting. Furthermore, to learn transferable information which may vary across features, we propose an adaptive transfer learning method that can detect and aggregate the feature-wise transferable structures. Non-asymptotic bound is provided for the estimation error of the target model, showing the robustness of the proposed method to high-dimensional covariate shifts.
