Standard Transformers Achieve the Minimax Rate in Nonparametric Regression withSmooth Targets

讲座题目：Standard Transformers Achieve the Minimax Rate in Nonparametric Regression withSmooth Targets

讲座时间 Datetime:2026年9月20日，周日，11：00 - 12：00

地点 Venue: 海琴2号A457

主持人 Host：杨云斐副教授

报告人 Speaker: 赖彦铭博士

单位 Affiliation:香港理工大学

报告摘要：

The tremendous success of Transformer models in fields such as large language models and computer vision necessitates a rigorous theoretical investigation. In this talk, we show that standard Transformers can approximate Hölder functionswith arbitrary precision. Building upon this approximation result, we demonstrate that standard Transformers achieve the minimax optimal rate in nonparametric regression for Hölder target functions. It is worth mentioning that, by introducing two metrics: the size tuple and the dimension vector, we provide a fine-grained characterization of Transformer structures, which facilitates future research on the generalization and optimization errors of Transformers with different structures. These findings provide theoretical justification for the powerful capabilities of Transformer models.

报告人简介：

赖彦铭博士是香港理工大学应用数学系的博士后研究员。此前，他在香港科技大学获得数学博士学位，师从汪扬教授。他的研究方向为神经网络的数学理论，其研究成果发表于Journal of Machine Learning Research, IEEE Transactions on Information Theory, SIAM Journal on Mathematical Analysis 和 International Conference on Machine Learning 等机器学习主流期刊和会议。