中國(guó)科學(xué)技術(shù)大學(xué)郭瀟作了一場(chǎng)題為“The effect of L1 penalization on condition number constrained estimation of precision matrix(L1處罰的精密矩陣條件數(shù)量的限制估計(jì)的影響)”的在職研究生講座。
中國(guó)科學(xué)技術(shù)大學(xué)有數(shù)學(xué)、物理學(xué)、化學(xué)、地球物理學(xué)、生物學(xué)、科學(xué)技術(shù)史、力學(xué)、儀器科學(xué)與技術(shù)、材料科學(xué)與工程、動(dòng)力工程及工程熱物理、電子科學(xué)與技術(shù)、信息與通信工程、控制科學(xué)與工程、計(jì)算機(jī)科學(xué)與技術(shù)、核科學(xué)與技術(shù)、環(huán)境科學(xué)與工程、生物醫(yī)學(xué)工程、管理科學(xué)與工程、工商管理、公共管理、軟件工程、安全科學(xué)與工程、統(tǒng)計(jì)學(xué)在職研究生、生態(tài)學(xué)、地質(zhì)學(xué)、天文學(xué)、哲學(xué)、應(yīng)用經(jīng)濟(jì)學(xué)、新聞傳播學(xué)、法學(xué)、大氣科學(xué)、光學(xué)工程一級(jí)學(xué)科碩士學(xué)位授權(quán)點(diǎn)。
大型精密矩陣估計(jì)是高維推論的基礎(chǔ)。一個(gè)重要的問題是要處理的精度矩陣的估計(jì),通常是有限的樣品中遇到的病態(tài),但在文獻(xiàn)中很少影響。在本文中,我們專注于通過征收界的估計(jì),從而有效地保證了精心調(diào)理的條件數(shù)估計(jì)精度矩陣。具體來說,我們建議基于相關(guān)的估計(jì),既條件數(shù)和L1懲罰約束,產(chǎn)生具有收斂理論上保證率精度矩陣估計(jì)。這一結(jié)果進(jìn)一步使我們能夠證明,結(jié)合了L1懲罰是必要的實(shí)現(xiàn)典型的高維設(shè)置所產(chǎn)生的估計(jì)的一致性,而當(dāng)L1罰不存在將出現(xiàn)不一致的情況;诔藬(shù)的交替方向法的算法開發(fā)來實(shí)現(xiàn)所提出的方法,揭示了模擬研究,表現(xiàn)令人滿意。該方法以呼叫中心數(shù)據(jù)的應(yīng)用說明。
原文:Estimation of large precision matrices is fundamental to high-dimensional inference. An important issue is to deal with ill-conditioning of the precision matrix estimate, typically encountered in finite-samples, but was rarely studied in the literature. In this paper, we focus on estimating the precision matrix by imposing a bound on the condition number of the estimate, which effectively ensures well-conditioning. Specifically, we propose a correlation-based estimator, constrained with both the condition number and the L1 penalty, yielding a precision matrix estimator with theoretically guaranteed rate of convergence. This result further enables us to demonstrate that incorporating the L1 penalty is necessary for achieving consistency of the resulting estimator in typical high-dimensional settings, while inconsistency will occur when the L1 penalty is absent. An algorithm based on the alternating direction method of multipliers is developed to implement the proposed method, which reveals the satisfactory performance in simulation studies. An application of the method to a call center data is illustrated.