報告題目：Scalable and Model-free Methods forMulticlass Probability Estimation
報 告 人：Helen Zhang教授美國亞利桑那大學
報告時間：2019年7月11日上午9:00-9:40
報告地點：數學樓一樓第一報告廳
報告摘要:
Classicalapproaches for multiclass probability estimation are mostly model-based, suchas logistic regression or LDA, by making certain assumptions on the underlyingdata distribution. We propose a new class of model-free methods to estimateclass probabilities based on large-margin classifiers. The method is scalablefor high-dimensional data by employing the divide-and-conquer technique, whichsolves multiple weighted large-margin classifiers and then constructsprobability estimates by aggregating multiple classification rules. Withoutrelying on any parametric assumption, the estimates are shown to be consistentasymptotically. Both simulated and real data examples are presented toillustrate performance of the new procedure.
報告人簡介：
Helen Zhang是美國亞利桑那大學數學系，統計學跨學科和應用數學跨學科的教授。她於2002年從美國威斯康星大學麥迪遜分校獲得博士學位，2002年至2011年期間擔任美國北卡羅來納州立大學的助理教授和副教授。她的主要研究方向包括非參數建模，統計機器學習，高維數據分析和應用的理論和方法。她目前是ISI journal Stat的主編。 她是美國統計學會和國際數學統計學會當選會員和2019年IMS Medallion lecture speaker。
報告題目：Learning ParameterHeterogeneity over Networks: A Distributed Tree-Based Fused-Lasso Approach
報 告 人：Zhengyuan Zhu教授美國愛荷華州立大學
報告時間：2019年7月11日上午9:40-10:20
報告地點：數學樓一樓第一報告廳
報告摘要:
We propose an adaptive fused-lasso based coefficient subgroupapproach in the decentralized network system. The major goal is to improve themodel estimation efficiency by aggregating the neighbors & information aswell as identify the subgroup membership for each node in the network. In particular,a tree-based $l_1$ penalty is proposed to save the computation andcommunication cost. We also design a decentralized generalized alternatingdirection method of multiplier algorithm for solving the objective function inparallel. The theoretical properties are derived to guarantee both the modelconsistency and the algorithm convergence. Thorough numerical experiments arealso conducted to back up our theory, which also show that our approachoutperforms in the aspects of the estimation accuracy, computation speed andcommunication cost.
報告人簡介：
Zhengyuan Zhu是美國愛荷華州立大學LAS院長統計學教授，也是調查統計與方法學中心主任。他於2002年獲得美國芝加哥大學統計學博士學位，並於2009年加入沒過過愛荷華州立大學，之前他擔任美國北卡羅來納大學教堂山分校統計學助理教授。他擁有空間統計，調查統計，空間抽樣設計和時間序列分析方面的專業知識，並對環境統計，遙感，自然資源調查和農業統計中的應用感興趣。他是許多國家大型縱向調查的PI和co-PI，包括美國國家資源調查，美國BLM管理土地調查和保護影響評估項目調查。
報告題目：Model averaging prediction for time series models witha divergingnumberof parameters
報 告 人：Guohua Zou教授首都師範大學
報告時間：2019年7月11日上午10:40-11:20
報告地點：數學樓一樓第一報告廳
報告摘要:
Animportant problem with model averaging approach is the choice ofweights.In this paper, a generalized Mallows model averaging (GMMA) criterion forchoosingweights is developed in the context of an infinite order autoregressive(AR(infinity))process. The GMMA method adapts to the circumstances in which thedimensionsof candidate models can be large and increase with the sample size. TheGMMAmethod is shown to be asymptotically optimal in the sense of obtaining thebestout-of-sample mean-squared prediction error (MSPE) for both the independent-realizationand the same-realization predictions, which, as a byproduct, solves aconjecture put forward by Hansen (2008) that the well-known Mallows modelaveraging (MMA) criterion from Hansen (2007) is asymptotically optimal forpredicting the future of a times series. The rate of the GMMA based weightestimator tending to the optimal weight vector minimizing theindependent-realization MSPE is derived as well. Both simulation experiment andreal data analysis illustrate the merits of GMMA method in the prediction ofAR(infinity) process.
報告人簡介：
Zou教授於1995年獲得中國科學院系統科學研究所統計學博士學位。他獲得國家傑出青年科學基金項目資助。 他的主要研究興趣包括利用統計理論和方法來分析實際的經濟，醫學和遺傳數據。他的研究領域包括統計模型選擇和平均，調查抽樣，統計決策理論和統計遺傳學。他特別關注的是混合效應模型，預測試估計和計量經濟學預測因子和測試的敏感性，估計量和預測因子的最優性，如可接受性和極小性，調查中的設計和數據分析，以及疾病和基因之間的聯繫和關聯研究。
報告題目：How many people canthe Earth support?
報 告 人：Joel E. Cohen教授美國洛克菲勒大學和哥倫比亞大學，美國科學院院士
報告時間：2019年7月11日下午2:00-3:00
報告地點：數學樓一樓第二報告廳
報告摘要:
Historical estimates of how many people the Earth can support ranged from<109to >1030people. The estimates had widely differentassumptions, methods, and purposes. To make "How many people can the Earthsupport?" into a scientifically meaningful question requires at least 11 basicassumptions. Even with clear assumptions, estimates of how many people theEarth can support depend on interactions of populations, environments,economies, and cultures. These interactions are complex and poorly understood. Henceestimates of how many people the Earth can support are highly uncertain.
Although no oneknows how many people the Earth can support, people can do three kinds ofthings to make life better now and in the future: create a bigger pie (createand use new technologies), bring fewer forks to the table (slow populationgrowth and reduce the material throughput of consumption), and practice bettermanners (resolve conflicts peacefully, trade more efficiently, and govern lessviolently and less corruptly). Universal basic and secondary education would helpfuture generations do all three.Educationshould give children a good understanding of the workings of their own bodiesand minds and the bodies and minds of others.
Butchildren can learn only if their brains andbodies work well. Hence universal education requires good nutrition for all childrenand their mothers. Despite a global abundance of food, nearly a quarter ofchildren under 5 years old are stunted from chronic undernutrition andinfection. The reduction in later economic output when children become workersdue to malnutrition in childhood is far greater than the cost of feeding allchildren and their mothers well. Feeding and educating all children and theirmothers well is profitable economically and desirable morally.
報告人簡介：
Joel E.Cohen是美國洛克菲勒大學的Abby Rockefeller Mauzé教授和哥倫比亞大學的教授。他和他的同事使用數學，統計和計算工具研究人口，生態系統和環境。 他的工作重點是影響人類健康的現象，人類與之相互作用的其他物種以及人類環境。最近的例子包括食物網，昆蟲傳播的感染，龍捲風和人口動態。Cohen使用模型來預測未來的人口增長，國際移民，生命以及教育與生育的相互作用。
Cohen教授在哈佛大學接受教育，並獲得了學士學位。以優異成績，獲得兩個碩士學位和兩個博士學位，一個是應用數學，另一個是人口科學和熱帶公共衛生。他一直在哈佛大學任教，直到1975年，他加入洛克菲勒大學，擔任教授和人口實驗室負責人。此外，自1995年以來，他一直是哥倫比亞大學國際和公共事務學院的教授。他還隸屬於哥倫比亞大學地球與環境科學系及統計系。他在芝加哥大學統計系有名譽任命。
他是美國國家科學院院士，美國藝術與科學學院院士，美國哲學學會會員。他獲得過麥克阿瑟基金會獎學金，這一被稱為「天才」的獎，以及古根海姆獎獎學金，日本科學促進會的獎學金。他分享了泰勒環境成就獎和華盛頓特區泛美衛生組織Fred L. Soper獎的成果，以供人們研究南美錐蟲病。他曾在阿根廷，中國，英國，法國和日本擔任名譽和訪問學術任命。Cohen教授獲得了人口委員會頒發的第一個Olivia Schieffelin Nordberg獎，以表彰他1995年出版的How Many People Can the Earth Support?他還編寫或編輯了其他13本書，其中包括兩本關於普及教育，一系列科學和數學笑話，絕對零重力以及430多篇科學論文和章節。