机器学习与统计学（MLSTAT）会议是由中国现场统计研究会机器学习分会主办的学术会议。会议旨在促进机器学习与统计领域的国内外学者进行学术交流，引领机器学习与统计共同交叉发展的学术文化，推动作为数据科学与人工智能的奠基性学科的进步，以此助力相关数字经济产业的发展。
第四届机器学习与统计学会议（MLSTAT2026）将于2026年7月15日-17日在西南财经大学（四川成都市）举行。会议将邀请20位左右青年学者就机器学习、人工智能、统计学和应用数学等相关领域的前沿进展做大会主题报告，同时欢迎在读博士生进行墙报展示。
主题：Norm-Based Generalization Under Scaling: The φ-Curve from Deterministic Equivalence to Function Spaces
报告人简介：刘方辉，上海交通大学自然科学研究院与数学科学学院副教授，研究方向为机器学习数学理论与大模型机理分析。其主要研究工作包括函数空间视角下的机器学习理论、尺度扩展下的泛化理论，并进一步推动其在大模型微调与参数高效训练中的应用。近五年发表论文在SIAM/JMLR/NeurIPS/ICML/ICLR/TPAMI 20余篇，相关研究获得国家自然科学基金委（海外优青项目）、英国皇家学会、阿兰·图灵研究所及谷歌等机构资助。2019 年博士毕业于上海交通大学，曾在 KU Leuven、EPFL 从事博士后研究，英国 University of Warwick担任助理教授。入选TUM 全球访问教授计划，获 AAAI 2024 新教师奖。担任 NeurIPS、ICLR、AISTATS等会议领域主席。
报告摘要：In this talk, I will discuss some fundamental questions in modern machine learning: What is a suitable model capacity of a modern machine learning model? How to precisely characterize the test risk under such a model capacity? What is the corresponding function space induced by such a model capacity? What are the fundamental limits of statistical/computational learning efficiency within space? My talk will partly answer the above questions, through the lens of norm-based capacity control. By deterministic equivalence, we provide a precise characterization of how the estimator's norm concentrates and how it governs the associated test risk. Our results show that the predicted learning curve admits a phase transition from under- to over-parameterization, but no double descent behavior, and reshapes scaling laws as well. Additionally, I will talk about the path-norm based capacities and the induced Barron spaces to understand the fundamental limits of statistical efficiency, particularly in terms of sample complexity and dimension dependence—highlighting key statistical-computational gaps. Talk based on https://arxiv.org/abs/2502.01585, https://arxiv.org/abs/2404.18769.
主题：Escaping Local Minima Deterministically and Provably in Matrix Sensing: Power of Simulated Over-Parameterization
报告人简介：马梓业现为香港城市大学计算机系助理教授（presidential assistant professor），博士毕业于加州大学伯克利分校（UC Berkeley）电子工程与计算机科学系。他的主要研究方向是AI基础算法，机器学习理论，以及数学优化（非凸优化）。他的论文主要发表在主流人工智能会议及期刊上（NeurIPS, ICML, JMLR, AISTATS, AAAI等），也曾多次获得口头报告（oral）荣誉，以及2023年的AISTATS Notable Paper。他曾获批香港研究资助局（rgc）的 "杰出青年学者计划（ecs）" 以及国自然青年科学基金（C类）。
报告摘要：Low-rank matrix sensing is a fundamental yet challenging nonconvex problem whose optimization landscape typically contains numerous spurious local minima, making it difficult for gradient-based optimizers to converge to the global optimum. Recent work has shown that tensor over-parameterization can, in principle, convert such local minima into strict saddle points; which also serves as a theoretical footnote to why scaling can improve generalization and performance in modern machine learning models like LLMs. Motivated by this observation, we propose a new escape mechanism that simulates the landscape and escape direction of the tensor-lifted space, without resorting to actually lifting the problem, since that would be computationally intractable. In essence, we designed a mathematical framework to project over-parametrized escape directions onto the original parameter space that could guarantee a strict decrease of objective value from existing local minima. To the best of our knowledge, this represents the first deterministic framework that could escape spurious local minima with guarantee, especially without using random perturbations or heuristic estimates.
主题：Contextual Linear Optimization under Full and Partial Feedback
报告人简介：毛小介，清华大学经济管理学院管理科学与工程系副教授，2016年获得武汉大学经济学学士学位，2021年获得美国康奈尔大学统计与数据科学专业博士学位。研究兴趣主要包括因果推断、数据驱动的决策方法、统计机器学习等。
报告摘要：This talk is about Contextual Linear Optimization (CLO) across two feedback regimes, where we study the traditional two-stage Estimate-Then-Optimize (ETO) approach and the new integrated Induced Empirical Risk Minimization (IERM) framework. In the full-feedback setting, we theoretically demonstrate that under model correct specification, ETO can surprisingly achieve faster regret convergence rates than IERM by leveraging problem-specific geometric properties. In partial-feedback settings (bandit and semi-bandit), we propose a unified offline IERM framework and establish novel fast-rate guarantees. Numerical experiments on shortest path problems validate our theoretical findings across different regimes.
会议注册信息:
为了确保会议顺利开展，本次会议将少量收取注册费，会务组将承担会议期间的用餐，其他费用敬请自理！
注册费用：学生代表 200元，其他代表500元。
报名截止时间：2026年6月30日。
会议信息将通过本公众号及会议网站及时更新，欢迎大家积极关注。会议注册是通过会议网站注册。会议注册网址：
https://ml-stat.github.io/MLSTAT2026/register/
联系方式:
邮箱：mlstat2026@126.com
电话：028-87092330
