Gaohang Yu
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
Gaohang Yu has a total of 90 co-authors. Here are the top co-authors by number of shared publications:
- Liqun Qi
Department of Applied Mathematics, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong
7 shared publications - Yisheng Song
School of Mathematical Sciences, Chongqing Normal University, Chongqing, China
6 shared publications Lutai Guan
Sun Yat-sen University
5 shared publications- Dalal Bardou
LMIA Laboratory, Department of Computer Science, Abbes Laghrour University of Khenchela, Khenchela, Algeria
4 shared publications - Laishui Lv
College of Information Engineering, China Jiliang University, Hangzhou, China
4 shared publications - Shanzhou Niu
Gannan Normal University
4 shared publications
Latest Publications
- 5 November 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 299
VLCA: Vision-language feature enhancement with cross-Attention learning for facial expression recognition
- 5 November 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 299
VLCA: Vision-language feature enhancement with cross-Attention learning for facial expression recognition
- 20 September 2025
- Published by Springer Nature
- p. 576-585
tCURLoRA: Tensor CUR Decomposition Based Low-Rank Parameter Adaptation and Its Application in Medical Image Segmentation
- 20 September 2025
- Published by Springer Nature
- p. 576-585
tCURLoRA: Tensor CUR Decomposition Based Low-Rank Parameter Adaptation and Its Application in Medical Image Segmentation
- 1 September 2025
- Published by Elsevier in Artificial Intelligence in Medicine
- Vol. 170, 103254
LoRA-PT: Low-rank adapting UNETR for hippocampus segmentation using principal tensor singular values and vectors
- 1 September 2025
- Published by Elsevier in Artificial Intelligence in Medicine
- Vol. 170, 103254
LoRA-PT: Low-rank adapting UNETR for hippocampus segmentation using principal tensor singular values and vectors
- 18 June 2025
- Published by Springer Nature in Applied Intelligence
- Vol. 55 (10) , 1-21
CLCFE: complementary loss coupling for feature-enhanced few-shot fine-grained visual recognition
- 18 June 2025
- Published by Springer Nature in Applied Intelligence
- Vol. 55 (10) , 1-21
CLCFE: complementary loss coupling for feature-enhanced few-shot fine-grained visual recognition
- 28 May 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (8) , 1-25
MERGE: multimodal-enhanced representation and guided ensemble for pneumonia recognition in chest X-ray images
- 28 May 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (8) , 1-25
MERGE: multimodal-enhanced representation and guided ensemble for pneumonia recognition in chest X-ray images
- Open Access
- 27 April 2025
- Published by International Academic Press in Statistics, Optimization & Information Computing
- Vol. 13 (6) , 2560-2574
Randomized Algorithms for Low-Rank Tensor Completion in TT-Format
Tensor completion is a crucial technique for filling in missing values in multi-dimensional data. It relies on the assumption that such datasets have intrinsic low-rank properties, leveraging this to reconstitute the dataset using low-rank decomposition or other strategies. Traditional approaches often lack computational efficiency,particularly with singular value decomposition (SVD) for large-scale tensor. Furthermore, fixed-rank SVD methods struggle with determining a suitable initial rank when data are incomplete. This paper introduces two novel randomized algorithms designed for low-rank tensor completion in tensor train (TT) format, named TTrandPI and FPTT. The TTrandPI algorithm integrates randomized tensor train (TT) decomposition with power iteration techniques, thereby enhancing computational efficiency and accuracy by improving spectral decay and minimizing tail energy build-up. Meanwhile, the FPTT algorithm utilizes a fixed-precision low-rank approximation approach that adaptively selects tensor ranks based on error tolerance levels, thus reducing the dependence on a predetermined rank. By conducting numerical experiments on synthetic data, color images, and video sequences, both algorithms exhibit superior performance compared to some existing methods.
- 27 April 2025
- Published by International Academic Press in Statistics, Optimization & Information Computing
- Vol. 13 (6) , 2560-2574
Open AccessRandomized Algorithms for Low-Rank Tensor Completion in TT-Format
Tensor completion is a crucial technique for filling in missing values in multi-dimensional data. It relies on the assumption that such datasets have intrinsic low-rank properties, leveraging this to reconstitute the dataset using low-rank decomposition or other strategies. Traditional approaches often lack computational efficiency,particularly with singular value decomposition (SVD) for large-scale tensor. Furthermore, fixed-rank SVD methods struggle with determining a suitable initial rank when data are incomplete. This paper introduces two novel randomized algorithms designed for low-rank tensor completion in tensor train (TT) format, named TTrandPI and FPTT. The TTrandPI algorithm integrates randomized tensor train (TT) decomposition with power iteration techniques, thereby enhancing computational efficiency and accuracy by improving spectral decay and minimizing tail energy build-up. Meanwhile, the FPTT algorithm utilizes a fixed-precision low-rank approximation approach that adaptively selects tensor ranks based on error tolerance levels, thus reducing the dependence on a predetermined rank. By conducting numerical experiments on synthetic data, color images, and video sequences, both algorithms exhibit superior performance compared to some existing methods.
- 8 January 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (2) , 1-37
Dara: distribution-aware representation alignment for semi-supervised domain adaptation in image classification
- 8 January 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (2) , 1-37
Dara: distribution-aware representation alignment for semi-supervised domain adaptation in image classification
- Open Access
- 7 January 2025
- Published by Institution of Engineering and Technology (IET) in IET Signal Processing
- Vol. 2025 (1)
Bayesian Robust Tensor Decomposition Based on MCMC Algorithm for Traffic Data Completion
Data loss is a common problem in intelligent transportation systems (ITSs). And the tensor‐based interpolation algorithm has obvious superiority in multidimensional data interpolation. In this paper, a Bayesian robust tensor decomposition method (MBRTF) based on the Markov chain Monte Carlo (MCMC) algorithm is proposed. The underlying low CANDECOMP/PARAFAC (CP) rank tensor captures the global information, and the sparse tensor captures local information (also regarded as anomalous data), which achieves a reliable prediction of missing terms. The low CP rank tensor is modeled by linear interrelationships among multiple latent factors, and the sparsity of the columns on the latent factors is achieved through a hierarchical prior approach, while the sparse tensor is modeled by a hierarchical view of the Student‐ t distribution. It is a challenge for traditional tensor‐based interpolation methods to maintain a stable performance under different missing rates and nonrandom missing (NM) scenarios. The MBRTF algorithm is an effective multiple interpolation algorithm that not only derives unbiased point estimates but also provides a robust method for the uncertainty measures of these missing values.
- 7 January 2025
- Published by Institution of Engineering and Technology (IET) in IET Signal Processing
- Vol. 2025 (1)
Open AccessBayesian Robust Tensor Decomposition Based on MCMC Algorithm for Traffic Data Completion
Data loss is a common problem in intelligent transportation systems (ITSs). And the tensor‐based interpolation algorithm has obvious superiority in multidimensional data interpolation. In this paper, a Bayesian robust tensor decomposition method (MBRTF) based on the Markov chain Monte Carlo (MCMC) algorithm is proposed. The underlying low CANDECOMP/PARAFAC (CP) rank tensor captures the global information, and the sparse tensor captures local information (also regarded as anomalous data), which achieves a reliable prediction of missing terms. The low CP rank tensor is modeled by linear interrelationships among multiple latent factors, and the sparsity of the columns on the latent factors is achieved through a hierarchical prior approach, while the sparse tensor is modeled by a hierarchical view of the Student‐ t distribution. It is a challenge for traditional tensor‐based interpolation methods to maintain a stable performance under different missing rates and nonrandom missing (NM) scenarios. The MBRTF algorithm is an effective multiple interpolation algorithm that not only derives unbiased point estimates but also provides a robust method for the uncertainty measures of these missing values.
- Open Access
- 18 February 2024
- Published by International Academic Press in Statistics, Optimization & Information Computing
- Vol. 12 (2) , 530-546
An Effective Randomized Algorithm for Hyperspectral Image Feature Extraction
Analyzing the spectral and spatial characteristics of Hyperspectral Imaging (HSI) in a three-dimensional space is a challenging task. Recently, there have been developments in 3D feature extraction methods based on tensor decomposition, which allow for the effective utilization of both global and local information in HSI. These methods also explore the inherent low-rank properties of HSI through tensor decomposition. In this paper, we propose a new approach called variable randomized T-product decomposition (Vrt-SVD), which is a variation of Tensor Singular Spectral Analysis. The goal of this approach is to improve the efficiency of tensor methods for feature extraction and reduce artifacts of image processing. By using a randomized algorithm based on the variable t-SVD, we are able to capture both global and local spatial and spectral information in HSI efficiently, which enables us to explore its low-rank characteristics. To evaluate the effectiveness of the extracted features, we use a Support Vector Machine (SVM) classifier to assess the accuracy of image classification. By conducting numerous numerical experiments, we provide strong evidence to show that the proposed method outperforms several advanced feature extraction techniques.
- 18 February 2024
- Published by International Academic Press in Statistics, Optimization & Information Computing
- Vol. 12 (2) , 530-546
Open AccessAn Effective Randomized Algorithm for Hyperspectral Image Feature Extraction
Analyzing the spectral and spatial characteristics of Hyperspectral Imaging (HSI) in a three-dimensional space is a challenging task. Recently, there have been developments in 3D feature extraction methods based on tensor decomposition, which allow for the effective utilization of both global and local information in HSI. These methods also explore the inherent low-rank properties of HSI through tensor decomposition. In this paper, we propose a new approach called variable randomized T-product decomposition (Vrt-SVD), which is a variation of Tensor Singular Spectral Analysis. The goal of this approach is to improve the efficiency of tensor methods for feature extraction and reduce artifacts of image processing. By using a randomized algorithm based on the variable t-SVD, we are able to capture both global and local spatial and spectral information in HSI efficiently, which enables us to explore its low-rank characteristics. To evaluate the effectiveness of the extracted features, we use a Support Vector Machine (SVM) classifier to assess the accuracy of image classification. By conducting numerous numerical experiments, we provide strong evidence to show that the proposed method outperforms several advanced feature extraction techniques.
- Open Access
Low-rank Tensor Train Decomposition Using TensorSketch
Tensor train decomposition is one of the most powerful approaches for processing high-dimensional data. For low-rank tensor train decomposition of large tensors, the alternating least squares (ALS) algorithm is widely used by updating each core tensor alternatively. However, it may suffer from the curse of dimensionality due to the large scale of subproblems. In this paper, a novel randomized proximal ALS algorithm is proposed for low-rank tensor train decomposition by using TensorSketch, which allows for efficient implementation via fast Fourier transform. The theoretical lower bounds of sketch size are estimated for approximating the optimal value of subproblems. Numerical experiments on synthetic and real-world data also demonstrate the effectiveness and efficiency of the proposed algorithm.
Open AccessLow-rank Tensor Train Decomposition Using TensorSketch
Tensor train decomposition is one of the most powerful approaches for processing high-dimensional data. For low-rank tensor train decomposition of large tensors, the alternating least squares (ALS) algorithm is widely used by updating each core tensor alternatively. However, it may suffer from the curse of dimensionality due to the large scale of subproblems. In this paper, a novel randomized proximal ALS algorithm is proposed for low-rank tensor train decomposition by using TensorSketch, which allows for efficient implementation via fast Fourier transform. The theoretical lower bounds of sketch size are estimated for approximating the optimal value of subproblems. Numerical experiments on synthetic and real-world data also demonstrate the effectiveness and efficiency of the proposed algorithm.