讲座一

Distributed Statistical Learning via Refitting Bootstrap Samples

主讲人

竺紫威

竺紫威博士，于2019年至2022年在密歇根大学安娜堡分校(UMich)担任统计学助理教授。在加入密歇根大学之前，他于2018年至2019年在剑桥大学统计实验室担任研究助理，这个实验室是由Richard J. Samworth教授主持的。他从美国普林斯顿大学运筹学与金融工程系获得博士学位，师从范剑青教授。他的主要研究方向为联邦/分布式统计学习、高维统计、稳健统计和缺失数据。

讲座简介

In this talk, I will introduce a one-shot distributed learning algorithm via refitting Bootstrap samples, which we refer to as ReBoot. Given the local models that are fit on multiple independent subsamples, ReBoot refits a new model on the union of the Bootstrap samples drawn from these local models. The whole procedure requires only one round of communication of model parameters. Theoretically, we analyze the statistical rate of ReBoot for generalized linear models (GLM) and noisy phase retrieval, which represent convex and non-convex problems respectively. In both cases, ReBoot provably achieves the full-sample statistical rate whenever the subsample size is not too small. In particular, we show that the systematic bias of ReBoot, i.e., the error that is independent of the number of subsamples, is O(n^-2) in GLM, where n is the subsample size. This rate is sharper than that of model parameter averaging and its variants, implying the higher tolerance of ReBoot with respect to data splits to maintain the full-sample rate. Simulation study demonstrates the statistical advantage of ReBoot over competing methods including averaging and CSL (Communication-efficient Surrogate Likelihood) with one round of gradient communication. Finally, we propose FedReBoot, an iterative version of ReBoot, to aggregate convolutional neural networks for image classification, which exhibits substantial superiority over FedAvg within early rounds of communication.

讲座时间

2023年4月14日（星期五）

下午2:00-3:00

讲座地点

柳林校区弘远楼408会议室

期数

澳门尼威斯人官网统计研究中心系列报告

讲座二

Multistate analysis of multitype recurrent event and failure time data with event feedbacks in biomarkers

主讲人

马绰欣

马绰欣博士，现为北京师范大学-香港浸会大学联合国际学院统计学副教授，博士毕业于英国曼彻斯特大学，期间主要从事多状态转移模型和纵向数据建模等方面的研究。加入北师港浸大之前，马绰欣博士在英国剑桥大学的生物统计中心和公共卫生系担任博士后研究助理，期间主要从事医学图像和基因数据统计建模的研究。目前感兴趣和正在开展的工作包括图像数据的张量回归，纵向数据的因果推断和多状态转移模型。马绰欣博士担任国际一流统计学期刊Biometrics, Biometrical Journal和Electronic Journal of Statistics以及国际知名医学杂志Frontiers in Public Health的审稿人，曾受邀参加英国皇家统计学会国际会议，泛华统计协会应用统计研讨会等国际大型学术会议，并在会上作特邀报告。马绰欣博士现在也是剑桥大学公共卫生系的访问研究员，与临床医学和统计学的研究人员保持着紧密的合作关系。

讲座简介

We propose a class of multistate models for the analysis of multitype recurrent event and failure time data when there are past event feedbacks in longitu-dinal biomarkers. It can well incorporate various effects, including time-dependent and time-independent effects, of different event paths or the number of occurrences of events of different types. Asymptotic unbiased estimating equations based on polynomial splines appro-ximation are developed. The consistency and asym-ptotic normality of the proposed estima-tors are provided. Simulation studies show that the naive estimators which either ignore the past event feedback or the measurement errors are biased. Our method has a better coverage probability of the time-varying/constant coefficients, compared to the naive methods. An appli-cation to the dataset from the Athero-sclerosis Risk in Communities Study, which is also the motivating example to develop the method, is presented.

讲座时间

2023年4月14日（星期五）

下午3:00-4:00

讲座地点

柳林校区弘远楼408会议室

期数

澳门尼威斯人官网统计研究中心系列报告

讲座三

A Synthetic Regression Model for Large Portfolio Allocation

主讲人

张文扬

张文扬教授是英国一流大学约克大学的统计学首席教授，统计学三大国际顶尖期刊之一 the Annals of Statistics 的副主编，商务和经济统计方面的国际顶尖期刊 Journal of Business & Economic Statistics 的副主编。张文扬教授主要从事大数据分析，金融数据分析，高维数据分析，非参数建模、时间序列分析、空间数据分析，多层次建模，生存分析，结构方程模型等方向的研究。他在国际顶尖学术期刊发表了很多非常有影响的学术论文，他关于ABC方法的一篇论文被引用超过3000多次。他曾先后在英国伦敦政治经济学院、英国 Kent 大学、英国 Bath 大学、英国 York 大学任教，现为英国 York 大学统计学首席教授。他曾是英国皇家统计学会科研委员会委员(历史上第三位华人担任该委员会委员)，曾经连续担任三届统计学三大国际顶尖期刊之一 Journal of the American Statistical Association 的副主编。

讲座简介

Portfolio allocation is an important topic in financial data analysis. In this talk, based on the mean-variance optimization principle, I will present a synthetic reg-ression model for construction of portfolio allocation, and an easy to implement approach to generate the synthetic sample for the model. Compared with the reg-ression approach in existing literature for portfolio allocation, the proposed method of generating the synthetic sample pro-vides more accurate approximation for the synthetic response variable when the number of assets under consideration is large. Due to the embedded leave-one-out idea, the synthetic sample generated by the proposed method has weaker within sample correlation, which makes the resulting portfolio allocation more close to the optimal one. I will show this intuitive conclusion is theoretically confirmed to be true by the asymptotic properties esta-blished. I will also show intensive simu-lation studies to compare the proposed method with the existing ones, and illustrate the proposed method works better. Finally, I will apply the proposed method to real data sets, and show very encouraging yielded returns.

讲座时间

2023年4月14日（星期五）

下午4:00-5:00

讲座地点

柳林校区弘远楼408会议室

期数

澳门尼威斯人官网统计研究中心系列报告