Bei Jiang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Bei Jiang has a total of 104 co-authors. Here are the top co-authors by number of shared publications:
- Linglong Kong
University of Alberta
19 shared publications - Matthew Pietrosanu
Department of Mathematical & Statistical Sciences, University of Alberta, Edmonton, Canada
6 shared publications - Chengkai Fan
Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta, Canada
5 shared publications - Wei Victor Liu
Department of Civil and Environmental Engineering and School of Mining and Petroleum Engineering, University of Alberta
5 shared publications - Na Zhang
Department of Mathematical and Statistical Sciences, University of Alberta, Edmonton, Canada
4 shared publications - Eva Petkova
Department of Population Health, Grossman School of Medicine, New York University, USA
3 shared publications
Latest Publications
- Open Access
- 18 May 2026
- Published by Wiley in Statistics in Medicine
- Vol. 45 (10-12) , e70582
Regularized Tensor Quantile Regression With Applications to Neuroimaging Data Analysis
This article proposes a regularized linear quantile regression model with a scalar response and tensor‐valued covariates. Our model uniquely regularizes the parameters of a low‐dimensional tensor effect decomposition through the tensor estimate rather than directly through the decomposition's parameters. We establish the computational and statistical properties of the proposed algorithm and estimators, both of which require separate treatment due to the quantile loss function. Simulation studies demonstrate the superiority of our model over existing tensor frameworks when traditional regression assumptions are violated. A real‐world neuroimaging analysis further highlights the interpretability benefits of our approach.
- 18 May 2026
- Published by Wiley in Statistics in Medicine
- Vol. 45 (10-12) , e70582
Open AccessRegularized Tensor Quantile Regression With Applications to Neuroimaging Data Analysis
This article proposes a regularized linear quantile regression model with a scalar response and tensor‐valued covariates. Our model uniquely regularizes the parameters of a low‐dimensional tensor effect decomposition through the tensor estimate rather than directly through the decomposition's parameters. We establish the computational and statistical properties of the proposed algorithm and estimators, both of which require separate treatment due to the quantile loss function. Simulation studies demonstrate the superiority of our model over existing tensor frameworks when traditional regression assumptions are violated. A real‐world neuroimaging analysis further highlights the interpretability benefits of our approach.
- Open Access
- 9 April 2026
- Published by Oxford University Press (OUP) in Biometrics
- Vol. 82 (2)
Personalized treatment design in the context of functional confounding
One of the primary goals of individualized treatment rule (ITR) methodology is to identify optimal decision rules using clinical predictors. While functional data has become increasingly available in biomedical research, there has been limited work on incorporating functional data into ITR estimation, particularly in observational studies. In this paper, we propose a novel approach that integrates outcome-weighted learning (OWL) with reproducing kernel Hilbert space to determine optimal treatment regimes involving functional data. Furthermore, to address the issue of data piling, we employ the distance-weighted discrimination classifier instead of traditional support vector machines. We establish the theoretical consistency of the decision functional estimator with its risk bound. Extensive simulations and the analysis of the Alzheimer’s Disease Neuroimaging Initiative dataset demonstrate the superior performance of our method compared to existing OWL approaches. The results highlight critical factors in Alzheimer’s Disease progression and reveal limitations of the original OWL method in this context.
- 9 April 2026
- Published by Oxford University Press (OUP) in Biometrics
- Vol. 82 (2)
Open AccessPersonalized treatment design in the context of functional confounding
One of the primary goals of individualized treatment rule (ITR) methodology is to identify optimal decision rules using clinical predictors. While functional data has become increasingly available in biomedical research, there has been limited work on incorporating functional data into ITR estimation, particularly in observational studies. In this paper, we propose a novel approach that integrates outcome-weighted learning (OWL) with reproducing kernel Hilbert space to determine optimal treatment regimes involving functional data. Furthermore, to address the issue of data piling, we employ the distance-weighted discrimination classifier instead of traditional support vector machines. We establish the theoretical consistency of the decision functional estimator with its risk bound. Extensive simulations and the analysis of the Alzheimer’s Disease Neuroimaging Initiative dataset demonstrate the superior performance of our method compared to existing OWL approaches. The results highlight critical factors in Alzheimer’s Disease progression and reveal limitations of the original OWL method in this context.
- 3 December 2025
- Published by Elsevier in Pattern Recognition
- Vol. 173
Fast online ℓ0 elastic net subspace clustering via a novel dictionary update strategy
- 3 December 2025
- Published by Elsevier in Pattern Recognition
- Vol. 173
Fast online ℓ0 elastic net subspace clustering via a novel dictionary update strategy
- Open Access
CBMA: Improving conformal prediction through Bayesian model averaging
Conformal prediction has emerged as a popular technique for facilitating valid predictive inference across a spectrum of machine learning models, under minimal assumption of exchangeability. Recently, Hoff (2023) showed that full conformal Bayes provides the most efficient prediction sets (smallest by expected volume) among all prediction sets that are valid at the $(1 - α)$ level if the model is correctly specified. However, a critical issue arises when the Bayesian model itself may be mis-specified, resulting in prediction set that might be suboptimal, even though it still enjoys the frequentist coverage guarantee. To address this limitation, we propose an innovative solution that combines Bayesian model averaging (BMA) with conformal prediction. This hybrid not only leverages the strengths of Bayesian conformal prediction but also introduces a layer of robustness through model averaging. Theoretically, we prove that the resulting prediction set will converge to the optimal level of efficiency, if the true model is included among the candidate models. This assurance of optimality, even under potential model uncertainty, provides a significant improvement over existing methods, ensuring more reliable and precise uncertainty quantification.Open AccessCBMA: Improving conformal prediction through Bayesian model averaging
Conformal prediction has emerged as a popular technique for facilitating valid predictive inference across a spectrum of machine learning models, under minimal assumption of exchangeability. Recently, Hoff (2023) showed that full conformal Bayes provides the most efficient prediction sets (smallest by expected volume) among all prediction sets that are valid at the $(1 - α)$ level if the model is correctly specified. However, a critical issue arises when the Bayesian model itself may be mis-specified, resulting in prediction set that might be suboptimal, even though it still enjoys the frequentist coverage guarantee. To address this limitation, we propose an innovative solution that combines Bayesian model averaging (BMA) with conformal prediction. This hybrid not only leverages the strengths of Bayesian conformal prediction but also introduces a layer of robustness through model averaging. Theoretically, we prove that the resulting prediction set will converge to the optimal level of efficiency, if the true model is included among the candidate models. This assurance of optimality, even under potential model uncertainty, provides a significant improvement over existing methods, ensuring more reliable and precise uncertainty quantification. - Open Access
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 2357-2365
Oblivious Johnson--Lindenstrauss embeddings for compressed Tucker decompositions
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 2357-2365
Open AccessOblivious Johnson--Lindenstrauss embeddings for compressed Tucker decompositions
- Open Access
- 3 October 2025
- Published by School of Statistics, Renmin University of China in Journal of Data Science
- p. 187-202
Differentially Private Bayesian Envelope Regression via Sufficient Statistic Perturbation
We propose a differentially private Bayesian framework for envelope regression, a technique that improves estimation efficiency by modelling the response as a function of a low-dimensional subspace of the predictors. Our method applies the analytic Gaussian mechanism to privatize sufficient statistics from the data, ensuring formal $(\epsilon ,\delta )$-differential privacy. We develop a tailored Gibbs sampling algorithm that performs valid Bayesian inference using only the noisy sufficient statistics. This approach leverages the envelope structure to isolate the variation in predictors that is relevant to the response, reducing estimation error compared to standard regression under the same privacy constraints. Through simulation studies, we demonstrate improved estimation accuracy and tighter credible intervals relative to a differentially private Bayesian linear regression baseline.
- 3 October 2025
- Published by School of Statistics, Renmin University of China in Journal of Data Science
- p. 187-202
Open AccessDifferentially Private Bayesian Envelope Regression via Sufficient Statistic Perturbation
We propose a differentially private Bayesian framework for envelope regression, a technique that improves estimation efficiency by modelling the response as a function of a low-dimensional subspace of the predictors. Our method applies the analytic Gaussian mechanism to privatize sufficient statistics from the data, ensuring formal $(\epsilon ,\delta )$-differential privacy. We develop a tailored Gibbs sampling algorithm that performs valid Bayesian inference using only the noisy sufficient statistics. This approach leverages the envelope structure to isolate the variation in predictors that is relevant to the response, reducing estimation error compared to standard regression under the same privacy constraints. Through simulation studies, we demonstrate improved estimation accuracy and tighter credible intervals relative to a differentially private Bayesian linear regression baseline.
- 24 September 2025
- Published by Springer Nature in Statistics and Computing
- Vol. 35 (6) , 1-14
Spilt conformal prediction with missing response
- 24 September 2025
- Published by Springer Nature in Statistics and Computing
- Vol. 35 (6) , 1-14
Spilt conformal prediction with missing response
- 22 April 2025
- Published by Springer Nature in Knowledge and Information Systems
- Vol. 67 (8) , 6553-6575
A Bayesian hierarchical model for orthogonal Tucker decomposition with oblivious tensor compression
- 22 April 2025
- Published by Springer Nature in Knowledge and Information Systems
- Vol. 67 (8) , 6553-6575
A Bayesian hierarchical model for orthogonal Tucker decomposition with oblivious tensor compression
- Open Access
Deep Fair Learning: A Unified Framework for Fine-tuning Representations with Sufficient Networks
Ensuring fairness in machine learning is a critical and challenging task, as biased data representations often lead to unfair predictions. To address this, we propose Deep Fair Learning, a framework that integrates nonlinear sufficient dimension reduction with deep learning to construct fair and informative representations. By introducing a novel penalty term during fine-tuning, our method enforces conditional independence between sensitive attributes and learned representations, addressing bias at its source while preserving predictive performance. Unlike prior methods, it supports diverse sensitive attributes, including continuous, discrete, binary, or multi-group types. Experiments on various types of data structure show that our approach achieves a superior balance between fairness and utility, significantly outperforming state-of-the-art baselines.
Open AccessDeep Fair Learning: A Unified Framework for Fine-tuning Representations with Sufficient Networks
Ensuring fairness in machine learning is a critical and challenging task, as biased data representations often lead to unfair predictions. To address this, we propose Deep Fair Learning, a framework that integrates nonlinear sufficient dimension reduction with deep learning to construct fair and informative representations. By introducing a novel penalty term during fine-tuning, our method enforces conditional independence between sensitive attributes and learned representations, addressing bias at its source while preserving predictive performance. Unlike prior methods, it supports diverse sensitive attributes, including continuous, discrete, binary, or multi-group types. Experiments on various types of data structure show that our approach achieves a superior balance between fairness and utility, significantly outperforming state-of-the-art baselines.
- 18 March 2025
- Published by Springer Nature in Mining, Metallurgy & Exploration
- Vol. 42 (2) , 665-684
Rapid Estimation of Truck Cycle Time in Open-Pit Mine Haulage Based on Feature-Optimized Machine Learning
- 18 March 2025
- Published by Springer Nature in Mining, Metallurgy & Exploration
- Vol. 42 (2) , 665-684
Rapid Estimation of Truck Cycle Time in Open-Pit Mine Haulage Based on Feature-Optimized Machine Learning