報(bào)告人：劉一鳴 副教授

報(bào)告題目：Identify the source of spikes: factor or mixture?

報(bào)告時(shí)間：2025年11月18日（周二）11:00-12:00

報(bào)告地點(diǎn)：云龍校區(qū)6號(hào)樓304會(huì)議室

主辦單位：數(shù)學(xué)與統(tǒng)計(jì)學(xué)院、數(shù)學(xué)研究院、科學(xué)技術(shù)研究院

報(bào)告人簡(jiǎn)介：

劉一鳴，暨南大學(xué)經(jīng)濟(jì)學(xué)院副教授。目前主要研究方向：機(jī)器學(xué)習(xí)、隨機(jī)矩陣、經(jīng)驗(yàn)似然及其相關(guān)應(yīng)用等。主持國(guó)自然科學(xué)基金，廣東省自然科學(xué)面上基金等項(xiàng)目。至今已在IEEE Transactions on Information Theory, Bernoulli, Statistica Sinica, Scandinavian Journal of Statistics等雜志發(fā)表論文15篇。

報(bào)告摘要：

We consider the problem of identifying the pattern of latent variables in high-dimensional linear latent variable models, which can also be interpreted as determining the source of spiked singular values in the data matrix. Specifically, we test whether the latent variables are continuous or categorical, a distinction which is crucial for data interpretation but challenging when the dimensionality is comparable to the sample size. To address this inference problem, we analyze the asymptotic behavior of empirical measures associated with singular vectors corresponding to large spiked singular values. Leveraging these insights, we propose a novel test statistic based on the eigenvector quantile differences and establish its theoretical performance under the null hypothesis. Simulation studies and real data analyses for breast cancer and glioblastoma gene expression datasets demonstrate the effectiveness and practical utility of our method.