研究生创新系列报告
报告时间：12月7日
报告地点：维格堂113
报告时间：09:00-09:40
报告题目：Parameter-transfer learning by Mallows model averaging with privacy protection
报告人：江芬 （中国科学技术大学）
报告摘要： Statistical modeling usually faces the challenges of insufficient samples and multi-source heterogeneity. Transfer learning provides an effective framework that leverages similar information from multiple sources. In this paper, we propose a parameter-transfer approach based on optimal model averaging with Mallows-type weight choice criterion. Compared with some existing two-stage transfer learning approaches, our approach does not require sharing multi-source samples but only utilizes parameter estimates of different models, which prevents the disclosure of individual data privacy. Theoretical analysis shows that our approach can achieve asymptotic optimality in prediction and weight convergence. Simulation studies and Genotype-Tissue Expression data application verify the effectiveness of our approach.
报告个介绍：江芬，来自于中国科学技术大学统计学专业，目前主要研究方向是模型平均和迁移学习。
报告时间：09:50-10:30
报告题目：Doubly robust estimation of optimal treatment regimes for survival data using an instrumental variable
报告人：夏俊文（中国人民大学）
报告摘要：In survival contexts, substantial literature exists on estimating optimal treatment regimes, where treatments are assigned based on personal characteristics to maximize the survival probability. These methods assume that a set of covariates is sufficient to deconfound the treatment-outcome relationship. However, this assumption can be limited in observational studies or randomized trials in which non-adherence occurs. Therefore, we propose a novel approach to estimating optimal treatment regimes when certain confounders are unobservable and a binary instrumental variable is available. Specifically, via a binary instrumental variable, we propose a semiparametric estimator for optimal treatment regimes by maximizing a Kaplan-Meier-like estimator of the survival function. Furthermore, to increase resistance to model misspecification, we construct novel doubly robust estimators. Since the estimators of the survival function are jagged, we incorporate kernel smoothing methods to improve performance. Under appropriate regularity conditions, the asymptotic properties are rigorously established. Moreover, the finite sample performance is evaluated through simulation studies. Finally, we illustrate our method using data from the National Cancer Institute's prostate, lung, colorectal, and ovarian cancer screening trial.
报告人介绍：夏俊文，中国人民大学统计学院统计学专业在读博士生，师从张景肖教授。主要研究方向为个性化最优决策，因果推断等。研究论文发表在Statistics and Computing上。
报告时间：12月8日
报告地点：维格堂113
报告时间：09:00-09:40
报告题目：Mixed membership network with the autoregressive structure
报告人：孙天一 （中国科学技术大学）
报告摘要：The study of community networks has received extensive attention, and the study of dynamic random block networks is one of the emerging directions of network analysis. Our work first proposes a mixed membership autoregressive network model, which gives the mixed-membership random block network a first-order autoregressive structure, and then discusses the problem of estimating community members. We build an AR-1 mixed spectral clustering algorithm on this model to estimate the membership matrix. In addition, we introduce an empirical eigenvalue threshold estimator to estimate the number of communities. Simulation results show that our method shows stronger generalization ability than previous methods for both random block community detection and mixed-membership community detection problems. Under different settings, the explicit error rate of our method does not exceed that of previous methods, and in many cases our method performs better. Application to real data proves the effectiveness of our method.
报告人介绍：孙天一，中国科学技术大学管理学院统计学专业在读博士生，主要研究方向为流行病传播模型，网络群落检测等。
报告时间：09:50-10:30
报告题目：Transfer Learning for Vector Autoregression
报告人：林余昌 （上海财经大学）
报告摘要: In this talk, we introduce a transfer learning procedure for high-dimensional time series data, using representation learning in the framework of vector autoregression. A three-stage regularized estimation approach is established with statistical convergence guarantee. The proposed method can improve the estimation and forecasting performance of the target dataset by borrowing similar information in representations from multiple source datasets. Notably our method allows for source datasets of different sequence lengths with unbalanced starting and ending points, which makes it much flexible to combine information from various datasets. Simulation studies and empirical analysis demonstrate the superiority of our transfer learning procedure over many single task learning methods, especially for short target datasets with large dimensions.
报告人介绍：林余昌，上海财经大学统计与管理学院在读博士生，主要研究方向为时间序列分析。研究论文曾发表在Journal of Time Series Analysis期刊上。
报告时间：10:40-11:20
报告题目：An Efficient Approach to Large Portfolio Optimization in Volatile Markets
报告人：苟梓康（中国人民大学）
报告摘要： With the global financial market experiencing continuous expansion and escalating volatility, the development of efficient strategies for large portfolio allocation has become critically important. In this work, we propose a novel Dantzig-type portfolio optimization (DPO) model designed to help investors navigate these challenges and optimize their portfolios effectively. The model separately incorporates L1 and nonconvex penalties, enabling the direct estimation of optimal portfolio weights while enforcing the sum constraint and accommodating both long and short positions. We establish the desired theoretical properties under mild regularity conditions, and introduce efficient parallel computing algorithms based on asset-splitting. Through extensive simulation studies, we investigate the superior effectiveness and efficiency of the DPO model and proposed algorithms. Furthermore, we illustrate the usefulness of the model by applying it to U.S. stock market datasets, including the S&P 500 and Russell 2000 indices.
报告人介绍：苟梓康，中国人民大学统计与大数据研究院2021级博士研究生，主要研究方向为高维数据分析，模型算法优化等。2023年阿里巴巴全球梦想家成员，曾多次荣获学业一等奖学金、优秀班团骨干等荣誉。
邀请人：马学俊