Xinmin Li
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
Research Subjects
Xinmin Li has a total of 54 co-authors. Here are the top co-authors by number of shared publications:
Hua Liang
4 shared publications- Hua Liang
Department of Statistics, George Washington University, Washington DC, USA
4 shared publications Juan Wang
School of Mathematics and Statistics, Qingdao University, Qingdao, People's Republic of China
4 shared publications- Chenchen Zou
School of Mathematics and Statistics, Qingdao University, Qingdao 266071, China
3 shared publications - Xiao Wang
Qingdao University
3 shared publications Jianxin Zhao
Basic Courses Department, Navy Submarine Academy, Qingdao, China
2 shared publications
Latest Publications
- Open Access
- 26 June 2026
- Published by Elsevier in SSRN Electronic Journal
RFF-CoRes: A Gaussian Process Framework with Explicit Random Fourier Feature and Corrective Residual for Forward and Inverse PDE Problems
We propose RFF-CoRes, a Gaussian-process framework for forward and inverse partial differential equations. The method represents the solution as an explicit Random Fourier feature mean plus a Gaussian-process residual correction. Instead of using an exact Gaussian kernel, we approximate the kernel by Random Fourier features, leading to a low-rank representation that enables Woodburyaccelerated posterior correction and avoids full kernel matrix factorization. This design retains the flexibility of Gaussian-process correction while substantially improving computational efficiency. The explicit RFF mean captures the dominant global and oscillatory solution structures, whereas the low-rank GP correction resolves local residual components. We further introduce a residualdriven adaptive refinement strategy that concentrates collocation points in regions with large PDE residuals, improving accuracy near boundaries, sharp gradients, and highly sensitive regions. Across elliptic, eikonal, convection–diffusion, Darcy flow, and lid-driven cavity benchmarks, RFF-CoRes consistently improves both accuracy and runtime efficiency.The results show that the proposed framework demonstrates strong versatility, robustness, and high accuracy in solving various types of forward and inverse partial differential equation problems, providing a scalable implementation path for building efficient and reliable physical information machine learning solvers.
- 26 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessRFF-CoRes: A Gaussian Process Framework with Explicit Random Fourier Feature and Corrective Residual for Forward and Inverse PDE Problems
We propose RFF-CoRes, a Gaussian-process framework for forward and inverse partial differential equations. The method represents the solution as an explicit Random Fourier feature mean plus a Gaussian-process residual correction. Instead of using an exact Gaussian kernel, we approximate the kernel by Random Fourier features, leading to a low-rank representation that enables Woodburyaccelerated posterior correction and avoids full kernel matrix factorization. This design retains the flexibility of Gaussian-process correction while substantially improving computational efficiency. The explicit RFF mean captures the dominant global and oscillatory solution structures, whereas the low-rank GP correction resolves local residual components. We further introduce a residualdriven adaptive refinement strategy that concentrates collocation points in regions with large PDE residuals, improving accuracy near boundaries, sharp gradients, and highly sensitive regions. Across elliptic, eikonal, convection–diffusion, Darcy flow, and lid-driven cavity benchmarks, RFF-CoRes consistently improves both accuracy and runtime efficiency.The results show that the proposed framework demonstrates strong versatility, robustness, and high accuracy in solving various types of forward and inverse partial differential equation problems, providing a scalable implementation path for building efficient and reliable physical information machine learning solvers.
- Open Access
A Penalized Orthogonal Kriging Method for Selecting a Global Trend
A kriging regression model is a popular and effective type of surrogate model in computer experiments. A significant challenge arises when the mean function of the model includes polynomial terms with unknown coefficients, leading to identifiability problems and potentially unreliable results. To overcome this problem, Plumlee and Joseph (2018) introduced an orthogonal kriging model. Variable selection for kriging models has been widely considered by researchers in computer experiments. In this paper, we introduce a new method for combining orthogonal kriging with penalized variable selection. Furthermore, an efficient algorithm is given to select the correct mean function. The simulation results and an example study with real data show that the proposed method is superior to others in variable recognition rate and prediction accuracy.
Open AccessA Penalized Orthogonal Kriging Method for Selecting a Global Trend
A kriging regression model is a popular and effective type of surrogate model in computer experiments. A significant challenge arises when the mean function of the model includes polynomial terms with unknown coefficients, leading to identifiability problems and potentially unreliable results. To overcome this problem, Plumlee and Joseph (2018) introduced an orthogonal kriging model. Variable selection for kriging models has been widely considered by researchers in computer experiments. In this paper, we introduce a new method for combining orthogonal kriging with penalized variable selection. Furthermore, an efficient algorithm is given to select the correct mean function. The simulation results and an example study with real data show that the proposed method is superior to others in variable recognition rate and prediction accuracy.
- 15 March 2025
- Published by Elsevier in Statistics & Probability Letters
- Vol. 222
Asymptotic optimality of generalized cross validation and regularized Mallows model averaging
- 15 March 2025
- Published by Elsevier in Statistics & Probability Letters
- Vol. 222
Asymptotic optimality of generalized cross validation and regularized Mallows model averaging
- Open Access
Fiducial Inference in Linear Mixed-Effects Models
We develop a novel framework for fiducial inference in linear mixed-effects (LME) models, with the standard deviation of random effects reformulated as coefficients. The exact fiducial density is derived as the equilibrium measure of a reversible Markov chain over the parameter space. The density is equivalent in form to a Bayesian LME with noninformative prior, while the underlying fiducial structure adds new benefits to unify the inference of random effects and all other parameters in a neat and simultaneous way. Our fiducial LME needs no additional tests or statistics for zero variance and is more suitable for small sample sizes. In simulation and empirical analysis, our confidence intervals (CIs) are comparable to those based on Bayesian and likelihood profiling methods. And our inference for the variance of random effects has competitive power with the likelihood ratio test.
Open AccessFiducial Inference in Linear Mixed-Effects Models
We develop a novel framework for fiducial inference in linear mixed-effects (LME) models, with the standard deviation of random effects reformulated as coefficients. The exact fiducial density is derived as the equilibrium measure of a reversible Markov chain over the parameter space. The density is equivalent in form to a Bayesian LME with noninformative prior, while the underlying fiducial structure adds new benefits to unify the inference of random effects and all other parameters in a neat and simultaneous way. Our fiducial LME needs no additional tests or statistics for zero variance and is more suitable for small sample sizes. In simulation and empirical analysis, our confidence intervals (CIs) are comparable to those based on Bayesian and likelihood profiling methods. And our inference for the variance of random effects has competitive power with the likelihood ratio test.
- 23 January 2025
- Published by Springer Nature in Computational Economics
- Vol. 66 (6) , 4493-4511
Nowcasting Monthly Chinese GDP with Mixed Frequency Data: A Model Averaging Approach
- 23 January 2025
- Published by Springer Nature in Computational Economics
- Vol. 66 (6) , 4493-4511
Nowcasting Monthly Chinese GDP with Mixed Frequency Data: A Model Averaging Approach
- 30 August 2024
- Published by Springer Nature in Statistics and Computing
- Vol. 34 (5) , 1-16
A comprehensive comparison of goodness-of-fit tests for logistic regression models
- 30 August 2024
- Published by Springer Nature in Statistics and Computing
- Vol. 34 (5) , 1-16
A comprehensive comparison of goodness-of-fit tests for logistic regression models
- 8 April 2024
- Published by Springer Nature in Journal of Systems Science and Complexity
- Vol. 37 (3) , 1080-1099
Optimal Model Average Prediction in Orthogonal Kriging Models
- 8 April 2024
- Published by Springer Nature in Journal of Systems Science and Complexity
- Vol. 37 (3) , 1080-1099
Optimal Model Average Prediction in Orthogonal Kriging Models
- 12 July 2023
- Published by Springer Nature in Statistics and Computing
- Vol. 33 (5) , 1-10
Use generalized linear models or generalized partially linear models?
- 12 July 2023
- Published by Springer Nature in Statistics and Computing
- Vol. 33 (5) , 1-10
Use generalized linear models or generalized partially linear models?
- Open Access
- 1 January 2023
- Published by Elsevier in SSRN Electronic Journal
Multi-Task Variable Selection in Gaussian Process
In this paper, we consider the variable selection in multi-output Gaussian process (GP) in a multi-task learning mode. The asymptotic selection consistency of multi-task Lasso is proved in the case of irregular regression matrices common in GP. And the overall consistency is unaffected by whether the inter-output correlations are modeled or not. Both the linear model of coregionalization (LMC) and the convolution model (CONV) can be reformed into and approximated by our framework. Simulation results show that the multi-task penalized GP displays higher F1 scores, less computation cost and competitive predictions compared with the single-task approach. And multi-task Lasso has higher computational efficiency and prediction accuracy than LMC and CONV.
- 1 January 2023
- Published by Elsevier in SSRN Electronic Journal
Open AccessMulti-Task Variable Selection in Gaussian Process
In this paper, we consider the variable selection in multi-output Gaussian process (GP) in a multi-task learning mode. The asymptotic selection consistency of multi-task Lasso is proved in the case of irregular regression matrices common in GP. And the overall consistency is unaffected by whether the inter-output correlations are modeled or not. Both the linear model of coregionalization (LMC) and the convolution model (CONV) can be reformed into and approximated by our framework. Simulation results show that the multi-task penalized GP displays higher F1 scores, less computation cost and competitive predictions compared with the single-task approach. And multi-task Lasso has higher computational efficiency and prediction accuracy than LMC and CONV.
- 2 July 2022
- Published by Taylor & Francis in Communications in Statistics - Simulation and Computation
- Vol. 53 (7) , 3418-3435
Confidence intervals for zero-inflated gamma distribution
The problems of constructing confidence intervals (CIs) for the mean of a gamma distribution containing zero values are considered. Three different methods for constructing such CIs based on Parametric Bootstrap (PB) method, fiducial inference and the method of variance of estimates recovery (MOVER) are developed. The performances of the proposed CIs are evaluated by Monte Carlo simulation and compared with a published method, asymptotic normal (AN) method. The simulation studies indicate that the proposed methods perform better than the existing method (AN) according to coverage probabilities (CPs), tail error rates, robustness and time comparison. All the methods are illustrated using a dataset for HIV risk behavior in adolescent girls and a dataset for the immunologic response to combination therapy for HIV children. Recommendations are made at the end of the articles.
- 2 July 2022
- Published by Taylor & Francis in Communications in Statistics - Simulation and Computation
- Vol. 53 (7) , 3418-3435
Confidence intervals for zero-inflated gamma distribution
The problems of constructing confidence intervals (CIs) for the mean of a gamma distribution containing zero values are considered. Three different methods for constructing such CIs based on Parametric Bootstrap (PB) method, fiducial inference and the method of variance of estimates recovery (MOVER) are developed. The performances of the proposed CIs are evaluated by Monte Carlo simulation and compared with a published method, asymptotic normal (AN) method. The simulation studies indicate that the proposed methods perform better than the existing method (AN) according to coverage probabilities (CPs), tail error rates, robustness and time comparison. All the methods are illustrated using a dataset for HIV risk behavior in adolescent girls and a dataset for the immunologic response to combination therapy for HIV children. Recommendations are made at the end of the articles.