Jun Fan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Jun Fan has a total of 66 co-authors. Here are the top co-authors by number of shared publications:
- Ding-Xuan Zhou
School of Mathematics and Statistics, University of Sydney, Sydney, Australia
9 shared publications - Linhao Song
School of Mathematical Science, Beihang University, Beijing, China
5 shared publications - Dao-Hong Xiang
School of Mathematical Sciences, Zhejiang Normal University, Zhejiang Key Laboratory of Intelligent Education Technology and Application, Jinhua, Zhejiang 321004, China
4 shared publications - Di-Rong Chen
School of Mathematical Science, Beihang University, Beijing, China
3 shared publications - Ming Yuan
Department of Statistics, Columbia University, New York, NY, USA
3 shared publications Ting Hu
School of Mathematics and Statistics, Wuhan University, Wuhan, Hubei 430072, P. R. China
3 shared publications
Latest Publications
- 30 October 2026
- Published by World Scientific Pub Co Pte Ltd in Analysis and Applications
- p. 1-43
Outcome weighted learning on data with low intrinsic dimension
Precision medicine aims to personalize treatment by identifying the optimal individualized treatment rule (ITR) that maximizes expected clinical outcomes. Outcome weighted learning (OWL) is an approach that has caught attention for estimating optimal ITR by leveraging each patient’s unique characteristics within a weighted classification framework. The curse of dimensionality in statistical learning significantly impacts the convergence rates of OWL algorithms, especially when analyzing high-dimensional biomedical data. We establish improved learning rates for OWL algorithms with varying Gaussian kernels, assuming the data lie on a low-dimensional intrinsic structure characterized by the box-counting dimension. Under noise or Besov smoothness assumptions, we derive a novel oracle inequality for OWL and obtain fast learning rates for the hinge loss function and [Formula: see text]-norm hinge loss function ([Formula: see text]), where the rates explicitly depend on the box-counting dimension of the prognostic variable domain. Additionally, we select the hyperparameters of OWL adaptively based on the data through a training-validation procedure. The same learning rates are obtained without assuming any prior knowledge of the data distribution.
- 30 October 2026
- Published by World Scientific Pub Co Pte Ltd in Analysis and Applications
- p. 1-43
Outcome weighted learning on data with low intrinsic dimension
Precision medicine aims to personalize treatment by identifying the optimal individualized treatment rule (ITR) that maximizes expected clinical outcomes. Outcome weighted learning (OWL) is an approach that has caught attention for estimating optimal ITR by leveraging each patient’s unique characteristics within a weighted classification framework. The curse of dimensionality in statistical learning significantly impacts the convergence rates of OWL algorithms, especially when analyzing high-dimensional biomedical data. We establish improved learning rates for OWL algorithms with varying Gaussian kernels, assuming the data lie on a low-dimensional intrinsic structure characterized by the box-counting dimension. Under noise or Besov smoothness assumptions, we derive a novel oracle inequality for OWL and obtain fast learning rates for the hinge loss function and [Formula: see text]-norm hinge loss function ([Formula: see text]), where the rates explicitly depend on the box-counting dimension of the prognostic variable domain. Additionally, we select the hyperparameters of OWL adaptively based on the data through a training-validation procedure. The same learning rates are obtained without assuming any prior knowledge of the data distribution.
- Open Access
- 28 September 2026
- Published by Springer Nature in Machine Learning
- Vol. 115 (10) , 232
Differentially Private Stochastic Gradient Descent for Outcome Weighted Learning
The development of individualized treatment rules in precision medicine seeks to optimize clinical outcomes for patients with varying responses to treatment. This paper focuses on outcome weighted learning, a method for estimating optimal treatment rules that take into account patient-specific characteristics within a weighted classification framework. We introduce a differentially private stochastic gradient descent algorithm within the framework of outcome weighted learning. This approach involves adding Gaussian noise to the gradient at each iteration, thereby ensuring the privacy of sensitive medical data while effectively managing large-scale datasets common in clinical practice. Unlike traditional differential privacy methods, which primarily focus on input-output type data, our approach integrates input-action-reward type data. This incorporation guarantees that both privacy and utility are preserved, with utility measured by the excess value function of the estimated individualized treatment rule. In our analysis, we establish convergence rates for the excess value function under logistic loss function, hinge loss function, and smoothed hinge loss function constructed via the Moreau envelope. These results provide rigorous utility guarantees for differentially private outcome weighted learning in precision medicine. Furthermore, empirical experiments support these theoretical findings and highlight the practical value of smoothing techniques in privacy-preserving learning for precision medicine.
- 28 September 2026
- Published by Springer Nature in Machine Learning
- Vol. 115 (10) , 232
Open AccessDifferentially Private Stochastic Gradient Descent for Outcome Weighted Learning
The development of individualized treatment rules in precision medicine seeks to optimize clinical outcomes for patients with varying responses to treatment. This paper focuses on outcome weighted learning, a method for estimating optimal treatment rules that take into account patient-specific characteristics within a weighted classification framework. We introduce a differentially private stochastic gradient descent algorithm within the framework of outcome weighted learning. This approach involves adding Gaussian noise to the gradient at each iteration, thereby ensuring the privacy of sensitive medical data while effectively managing large-scale datasets common in clinical practice. Unlike traditional differential privacy methods, which primarily focus on input-output type data, our approach integrates input-action-reward type data. This incorporation guarantees that both privacy and utility are preserved, with utility measured by the excess value function of the estimated individualized treatment rule. In our analysis, we establish convergence rates for the excess value function under logistic loss function, hinge loss function, and smoothed hinge loss function constructed via the Moreau envelope. These results provide rigorous utility guarantees for differentially private outcome weighted learning in precision medicine. Furthermore, empirical experiments support these theoretical findings and highlight the practical value of smoothing techniques in privacy-preserving learning for precision medicine.
- Open Access
- 10 September 2026
- Published by Elsevier in Journal of Approximation Theory
- Vol. 320
Classification with shallow neural networks
- 10 September 2026
- Published by Elsevier in Journal of Approximation Theory
- Vol. 320
Open AccessClassification with shallow neural networks
- 13 June 2026
- Published by World Scientific Pub Co Pte Ltd in Analysis and Applications
- p. 1-29
Fully online functional learning with streaming data
Analyzing large-scale functional data poses significant computational challenges due to high costs and substantial data storage needs. Additionally, traditional batch learning algorithms are not well-equipped to manage streaming data effectively. To address these issues, we propose a fully online learning algorithm designed for functional linear regression, which models the linear relationship between a scalar response and a functional predictor. Our approach employs Tikhonov regularization schemes within the framework of reproducing kernel Hilbert spaces (RKHS). A key feature of this fully online algorithm is its polynomially decaying regularization parameter, which adapts dynamically at each learning step, distinguishing it from the partially online algorithm that uses a fixed parameter. Within the functional linear model framework, we establish sufficient conditions for the convergence of the fully online algorithm in the RKHS norm. Additionally, we employ a capacity-independent approach to derive error bounds and almost sure convergence rates for both prediction and estimation, achieved through careful selection of step sizes and regularization parameters.
- 13 June 2026
- Published by World Scientific Pub Co Pte Ltd in Analysis and Applications
- p. 1-29
Fully online functional learning with streaming data
Analyzing large-scale functional data poses significant computational challenges due to high costs and substantial data storage needs. Additionally, traditional batch learning algorithms are not well-equipped to manage streaming data effectively. To address these issues, we propose a fully online learning algorithm designed for functional linear regression, which models the linear relationship between a scalar response and a functional predictor. Our approach employs Tikhonov regularization schemes within the framework of reproducing kernel Hilbert spaces (RKHS). A key feature of this fully online algorithm is its polynomially decaying regularization parameter, which adapts dynamically at each learning step, distinguishing it from the partially online algorithm that uses a fixed parameter. Within the functional linear model framework, we establish sufficient conditions for the convergence of the fully online algorithm in the RKHS norm. Additionally, we employ a capacity-independent approach to derive error bounds and almost sure convergence rates for both prediction and estimation, achieved through careful selection of step sizes and regularization parameters.
- 1 June 2026
- Published by Springer Nature in Applied Mathematics-A Journal of Chinese Universities
- Vol. 41 (2) , 480-508
Online outcome weighted learning with varying Gaussians and non-identical distributions
- 1 June 2026
- Published by Springer Nature in Applied Mathematics-A Journal of Chinese Universities
- Vol. 41 (2) , 480-508
Online outcome weighted learning with varying Gaussians and non-identical distributions
- 26 March 2026
- Published by Elsevier in Applied and Computational Harmonic Analysis
- Vol. 84
Learning Fréchet differentiable operators via prespecified neural operators
- 26 March 2026
- Published by Elsevier in Applied and Computational Harmonic Analysis
- Vol. 84
Learning Fréchet differentiable operators via prespecified neural operators
- 16 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Theory
- Vol. 70 (9) , 6547-6571
Distributed Gradient Descent for Functional Learning
In recent years, different types of distributed and parallel learning schemes have received increasing attention for their strong advantages in handling large-scale data information. In the information era, to face the big data challenges that stem from functional data analysis very recently, we propose a novel distributed gradient descent functional learning (DGDFL) algorithm to tackle functional data across numerous local machines (processors) in the framework of reproducing kernel Hilbert space. Based on integral operator approaches, we provide the first theoretical understanding of the DGDFL algorithm in many different aspects of the literature. On the way of understanding DGDFL, firstly, a data-based gradient descent functional learning (GDFL) algorithm associated with a single-machine model is proposed and comprehensively studied. Under mild conditions, confidence-based optimal learning rates of DGDFL are obtained without the saturation boundary on the regularity index suffered in previous works in functional regression. We further provide a semi-supervised DGDFL approach to weaken the restriction on the maximal number of local machines to ensure optimal rates. To our best knowledge, the DGDFL provides the first divide-and-conquer iterative training approach to functional learning based on data samples of intrinsically infinite-dimensional random functions (functional covariates) and enriches the methodologies for functional data analysis.
- 16 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Theory
- Vol. 70 (9) , 6547-6571
Distributed Gradient Descent for Functional Learning
In recent years, different types of distributed and parallel learning schemes have received increasing attention for their strong advantages in handling large-scale data information. In the information era, to face the big data challenges that stem from functional data analysis very recently, we propose a novel distributed gradient descent functional learning (DGDFL) algorithm to tackle functional data across numerous local machines (processors) in the framework of reproducing kernel Hilbert space. Based on integral operator approaches, we provide the first theoretical understanding of the DGDFL algorithm in many different aspects of the literature. On the way of understanding DGDFL, firstly, a data-based gradient descent functional learning (GDFL) algorithm associated with a single-machine model is proposed and comprehensively studied. Under mild conditions, confidence-based optimal learning rates of DGDFL are obtained without the saturation boundary on the regularity index suffered in previous works in functional regression. We further provide a semi-supervised DGDFL approach to weaken the restriction on the maximal number of local machines to ensure optimal rates. To our best knowledge, the DGDFL provides the first divide-and-conquer iterative training approach to functional learning based on data samples of intrinsically infinite-dimensional random functions (functional covariates) and enriches the methodologies for functional data analysis.
- 30 April 2024
- Published by Elsevier in Journal of Complexity
- Vol. 84
On the convergence of gradient descent for robust functional linear regression
- 30 April 2024
- Published by Elsevier in Journal of Complexity
- Vol. 84
On the convergence of gradient descent for robust functional linear regression
- 21 March 2024
- Published by MIT Press in Neural Computation
- Vol. 36 (4) , 718-743
Learning Korobov Functions by Correntropy and Convolutional Neural Networks
Combining information-theoretic learning with deep learning has gained significant attention in recent years, as it offers a promising approach to tackle the challenges posed by big data. However, the theoretical understanding of convolutional structures, which are vital to many structured deep learning models, remains incomplete. To partially bridge this gap, this letter aims to develop generalization analysis for deep convolutional neural network (CNN) algorithms using learning theory. Specifically, we focus on investigating robust regression using correntropy-induced loss functions derived from information-theoretic learning. Our analysis demonstrates an explicit convergence rate for deep CNN-based robust regression algorithms when the target function resides in the Korobov space. This study sheds light on the theoretical underpinnings of CNNs and provides a framework for understanding their performance and limitations.
- 21 March 2024
- Published by MIT Press in Neural Computation
- Vol. 36 (4) , 718-743
Learning Korobov Functions by Correntropy and Convolutional Neural Networks
Combining information-theoretic learning with deep learning has gained significant attention in recent years, as it offers a promising approach to tackle the challenges posed by big data. However, the theoretical understanding of convolutional structures, which are vital to many structured deep learning models, remains incomplete. To partially bridge this gap, this letter aims to develop generalization analysis for deep convolutional neural network (CNN) algorithms using learning theory. Specifically, we focus on investigating robust regression using correntropy-induced loss functions derived from information-theoretic learning. Our analysis demonstrates an explicit convergence rate for deep CNN-based robust regression algorithms when the target function resides in the Korobov space. This study sheds light on the theoretical underpinnings of CNNs and provides a framework for understanding their performance and limitations.
- 5 March 2024
- Published by Elsevier in Neural Networks
- Vol. 174, 106226
Generalization analysis of deep CNNs under maximum correntropy criterion
- 5 March 2024
- Published by Elsevier in Neural Networks
- Vol. 174, 106226
Generalization analysis of deep CNNs under maximum correntropy criterion