Research Interest（研究兴趣）:
Fundamental problems in machine learning and computer vision, especially including:
Meta-learning
Variational bayesian methods on inverse problems
Robust and interpretable deep learning
Suggested to Read (推荐阅读):
Understanding meta learning methodology as learning an explicit hyperparameter prediction policy shared by various training tasks:
[0] Jun Shu, Deyu Meng, Zongben Xu. Learning an Explicit Hyperparameter Prediction Policy Conditioned on Tasks. Journal of Machine Learning Research, 2023.
Meta-Learning-Rate-Schecule Net: Jun Shu, Yanwen Zhu, Qian Zhao, Deyu Meng, and Zongben Xu. MLR-SNet: Transferable LR Schedules for Heterogeneous Tasks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022. [code][supplementary material]
A new filter parametrization strategy, which not only finely represents 2D filters with zero error when the filter is not rotated, but also largely alleviates the fence-effect-caused quality degradation when the filter is rotated:
[0] Qi Xie, Qian Zhao, Zongben Xu and Deyu Meng. Fourier Series Expansion Based Filter Parametrization for Equivariant Convolutions. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022. [code][supplementary material]
On modeling non-iid noise with temporal noise prior:
[1] Hongwei Yong, Deyu Meng*, Wangmeng Zuo, Lei Zhang. Robust Online Matrix Factorization for Dynamic Background Subtraction, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017. [arxiv version][code][supplementary material]
Non-i.i.d. spectral noise modeling: Yang Chen, Xiangyong Cao, Qian Zhao, Deyu Meng*, Zongben Xu. Denoising Hyperspectral Image with Non-i.i.d. Noise Structure. IEEE Transactions on Cybernetics. 2017. [arxiv version][Appendix][code]
Non-i.i.d. spatial noise modeling: Xiangyong Cao, Qian Zhao, Deyu Meng*, Yang Chen, Zongben Xu. Robust Low-rank Matrix Factorization under General Mixture Noise Distributions, IEEE Transactions on Image Processing, 2016.[arxiv version][Code][Application on small target detection on infrared images][ICCVConferenceVersion(oral)][supplementary material]
On modeling non-iid noise in multi-view learning:
[2] Zongsheng Yue, Hongwei Yong, Deyu Meng*, Qian Zhao, Yee Leung, Lei Zhang. Robust Multi-view Subspace Learning with Non-independently and Non-identically Distributed Complex Noise. IEEE Transactions on Neural Networks and Learning Systems, 2019. [supplementary file] [code]
Application to hyper-spectral image restoration: Zongsheng Yue, Deyu Meng, Yanqing Sun, Qian Zhao. Hyperspectral Image Restoration under Complex Multi-Band Noises. Remote Sensing 10 (10), 1631, 2018.
On modeling noise for CT images:
[3] Qi Xie, Dong Zeng, Qian Zhao, Deyu Meng*, Zongben Xu, Jianhua Ma*, Zhenrong Liang, Robust Low-dose CT Sinogram Prepocessing via Exploiting Noise-generating Mechanism, IEEE Transactions on Medical Imaging, 2017.[code][github link][supplementary material]
On modeling noise for Lesion Detection from Fundus Images:
[4] Renzhen Wang, Benzhi Chen, Deyu Meng*, Lisheng Wang. Weakly-Supervised Lesion Detection from Fundus Images, IEEE Transactions on Medical Imaging, 2018.[Demo code]
On theoretical understandings for self-paced learning:
[5] Deyu Meng*, Qian Zhao, Lu Jiang. A Theoretical Understanding of Self-paced Learning. Information Sciences, 414: 319-328, 2017. [arxiv version][slides].
Convergence theory of SPL: Zilu Ma, Shiqi Liu, Deyu Meng*. On Convergence Property of Implicit Self-paced Objective. Information Sciences, 462, 132-140, 2018 [arxiv version].
Concave conjugacy theory of SPL: Shiqi Liu, Zilu Ma, Deyu Meng*. Understanding Self-Paced Learning under Concave Conjugacy Theory. Communications in Information and Systems, 18(1), 1-35, 2018 [arxiv version]
On applications of self-paced learning to multi-view/modality/feature problems:
[6] Fan Ma, Deyu Meng*, Xuanyi Dong, Yi Yang. Self-paced multi-view co-training. Journal of Machine Learning Research. 2020. [code][ICMLonference version][supplementary material][Application on few shot object detection][Application on weakly supervised object detection][Top 4 on TGSS competition][明报报道][文汇报报道][新浪报道][大公网报道][ChinaDaily报道][东方头条报道][成报报道][港校直通车报道]
[7]Xuanyi Dong, Liang Zheng, Fan Ma, Yi Yang, Deyu Meng*. Few-Example Object Detection with Model Communication. accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018.[arxiv version][code][dataset]
On automatical weight function learning for different training data bias cases:
[8] Jun Shu, Qi Xie, Lixuan Yi, Qian Zhao, Sanping Zhou, Zongben Xu, Deyu Meng*. Meta-Weight-Net: Learning an Explicit Mapping For Sample Weighting. NeurIPS, 2019.[code](Both sampling weighting schemes like easy sample emphasizing, e.g., self-paced learning, or hard sample emphasizing, e.g., focal loss, can be seen as special cases of this adaptive meta-learning framework.)
On blind image denoising by a variational denoising network:
[9] Zongsheng Yue,Hongwei Yong, Qian Zhao, Lei Zhang, Deyu Meng*. Variational Denoising Network: Toward Blind Noise Modeling and Removal. NeurIPS, 2019. [code](Our method can learn an approximate posterior to the true posterior with the latent variables including clean image and noise distribution (non-i.i.d.) conditioned on the input noisy image. Using this variational posterior expression, both tasks of blind image denoising and noise estimation can be naturally attained in a unique Bayesian framework.)
Learn A Noise Generator Simulating to Generate Real Image Noise: Zongsheng Yue, Qian Zhao, Lei Zhang, Deyu Meng*. Dual Adversarial Network: Toward Real Noise Removal and Noise Generation. ECCV, 2020.[arxiv][code][supplementary material]
On a new tensor sparsity measure based on Kronecker basis representation:
[10] Qi Xie, Qian Zhao, Deyu Meng*, Zongben Xu. Kronecker-Basis-Representation Based Tensor Sparsity and Its Applications to Tensor Recovery, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017. [code for MSI denoising][github link][code for MSI TC and RPCA][github link][conference version][ConferenceVersionCode][Application on CT reconstruction]
On an interpretable deep network for MS/HS fusion:
[11] Qi Xie, Minghao Zhou, Qian Zhao, Zongben Xu, Deyu Meng*. MHF-Net: An Interpretable Deep Network for Multispectral and Hyperspectral Image Fusion. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020. [code]
On variational Bayes' method in infinite-dimensional space, which makes the method feasible in general inverse problems like PDEs:
[12] Junxiong Jia, Qian Zhao, Zongben Xu, Deyu Meng*, Yee Leung. Variational Bayes' method for functions with applications to some inverse problems. SIAM Journal on Scientific Computing, 2020. [supplemental file][arxiv version]