Deng-Bao WANG  王登豹

Ph.D. Student

School of Computer Science and Engineering

Southeast University, China

Computer Building, Jiulonghu Campus of Southeast University, Nanjing, China
Email: wangdb[at]seu.edu.cn  OR  wangdb.seu[at]gmail.com

I am now a Ph.D. student (supervised by Prof. Min-Ling Zhang) of School of Computer Science and Engineering in Southeast University and a member of PALM group. I received B.Sc. degree from School of Computer and Control Engineering at Yantai University in 2016 and M.Sc. degree from College of Computer and Information Science at Southwest University in 2019.

Research Interests

• Uncertainty Calibration

• Weakly Supervised Learning

• Deep Learning Phenomena

Publications (†Equal Contribution)

• Distilling Reliable Knowledge for Instance-Dependent Partial Label Learning.
  D.-D. Wu†, D.-B. Wang†, M.-L. Zhang.
  In: Proceedings of the 38th AAAI Conference on Artificial Intelligence, Vancouver, Canada.
  
   AAAI 2024    Abstract    Paper    Appendix    Code    Bibtex  
  
  Partial label learning (PLL) refers to the classification task where each training instance is ambiguously annotated with a set of candidate labels. Despite substantial advancements in tackling this challenge, limited attention has been devoted to a more specific and realistic setting, denoted as instancedependent partial label learning (IDPLL). Within this contex, the assignment of partial labels depends on the distinct features of individual instances, rather than being random. In this paper, we initiate an exploration into a self-distillation framework for this problem, driven by the proven effectiveness and stability of this framework. Nonetheless, a crucial shortfall is identified: the foundational assumption central to IDPLL, involving what we term as partial label knowledge stipulating that candidate labels should exhibit superior confidence compared to non-candidates, is not fully upheld within the distillation process. To address this challenge, we introduce DIRK, a novel distillation approach that leverages a rectification process to DIstill Reliable Knowledge, while concurrently preserves informative fine-grained label confidence. In addition, to harness the rectified confidence to its fullest potential, we propose a knowledge-based representation refinement module, seamlessly integrated into the DIRK framework. This module effectively transmits the essence of similarity knowledge from the label space to the feature space, thereby amplifying representation learning and subsequently engendering marked improvements in model performance. Experiments and analysis on multiple datasets validate the rationality and superiority of our proposed approach. @inproceedings{AAAI24Wu,
  author = {Dong-Dong Wu and Deng-Bao Wang and Min-Ling Zhang},
  title = {Distilling Reliable Knowledge for Instance-Dependent Partial Label Learning},
  booktitle = {Proceedings of the 38th AAAI Conference on Artificial Intelligence},
  year = {2024}
  }


• On the Pitfall of Mixup for Uncertainty Calibration.
  D.-B. Wang, L. Li, P. Zhao, P.-A. Heng, M.-L. Zhang.
  In: Proceedings of the 34th IEEE/CVF Conference on Computer Vision and Pattern Recognition, Vancouver, Canada.
  
   CVPR 2023    Abstract    Paper    Appendix    Code    Bibtex  
  
  By simply taking convex combinations between pairs of samples and their labels, mixup training has been shown to easily improve predictive accuracy. It has been recently found that models trained with mixup also perform well on uncertainty calibration. However, in this study, we found that mixup training usually makes models less calibratable than vanilla empirical risk minimization, which means that it would harm uncertainty estimation when post-hoc calibration is considered. By decomposing the mixup process into data transformation and random perturbation, we suggest that the confidence penalty nature of the data transformation is the reason of calibration degradation. To mitigate this problem, we first investigate the mixup inference strategy and found that despite it improves calibration on mixup, this ensemble-like strategy does not necessarily outperform simple ensemble. Then, we propose a general strategy named mixup inference in training, which adopts a simple decoupling principle for recovering the outputs of raw samples at the end of forward network pass. By embedding the mixup inference, models can be learned from the original one-hot labels and hence avoid the negative impact of confidence penalty. Our experiments show this strategy properly solves mixup’s calibration issue without sacrificing the predictive performance, while even improves accuracy than vanilla mixup. @inproceedings{CVPR23Wang,
  author = {Deng-Bao Wang, Lanqing Li, Peilin Zhao, Pheng-Ann Heng, Min-Ling Zhang},
  title = {On the Pitfall of Mixup for Uncertainty Calibration},
  booktitle = {Proceedings of the 34th IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year = {2023}
  }


• Multiple-instance learning from triplet comparison bags.
  S. Shu, D.-B. Wang, S. Yuan, H. Wei, J. Jiang, L. Feng, M.-L. Zhang.
  ACM Transactions on Knowledge Discovery from Data, 2023.
  
   ACM TKDD 2023    Abstract    Paper    Code    Bibtex  
  
  Multiple-instance learning (MIL) solves the problem where training instances are grouped in bags, and a binary (positive or negative) label is provided for each bag. Most of the existing MIL studies need fully labeled bags for training an effective classifier, while it could be quite hard to collect such data in many real-world scenarios, due to the high cost of data labeling process. Fortunately, unlike fully labeled data, triplet comparison data can be collected in a more accurate and human-friendly way. Therefore, in this paper, we for the first time investigate MIL from only triplet comparison bags, where a triplet (𝑋𝑎, 𝑋𝑏, 𝑋𝑐 ) contains the weak supervision information that bag 𝑋𝑎 is more similar to 𝑋𝑏 than to 𝑋𝑐 . To solve this problem, we propose to train a bag-level classifier by the empirical risk minimization framework and theoretically provide a generalization error bound. We also show that a convex formulation can be obtained only when specific convex binary losses such as the square loss and the double hinge loss are used. Extensive experiments validate that our proposed method significantly outperforms other baselines


• Learning from Noisy Labels via Dynamic Loss Thresholding.
  H. Yang, Y.-Z. Jin, Z.-Y. Li, D.-B. Wang, X. Geng, M.-L. Zhang.
  IEEE Transactions on Knowledge and Data Engineering, 2023.
  
   IEEE TKDE 2023    Abstract    Paper    Code    Bibtex  
  
  Numerous researches have proved that deep neural networks (DNNs) can fit almost everything even given data with noisy labels, and result in poor generalization performance. However, recent studies suggest that DNNs tend to gradually memorize the data, moving from correct data to mislabeled data. Inspired by this finding, we propose a novel method named Dynamic Loss Thresholding (DLT). During the training process, DLT records the loss value of each sample and calculates dynamic loss thresholds. Specifically, DLT compares the loss value of each sample with the current loss threshold. Samples with smaller losses can be considered as clean samples with higher probability and vice versa. Then, DLT discards the potentially corrupted labels and further leverages self-training semi-supervised learning techniques. Experiments on CIFAR-10/100, WebVision and Clothing1M demonstrate substantial improvements over recent state-of-the-art methods. In addition, we investigate two real-world problems. Firstly, we propose a novel approach to estimate the noise rates of datasets based on the loss difference between the early and late training stages of DNNs. Secondly, we explore the effect of hard samples (which are difficult to be distinguished) on the process of learning from noisy labels. @inproceedings{TKDE23Yang,
  author = {Hao Yang, You-Zhi Jin, Zi-Yin Li, Deng-Bao Wang, Xin Geng, Min-Ling Zhang},
  title = {Learning from Noisy Labels via Dynamic Loss Thresholding},
  booktitle = {IEEE Transactions on Knowledge and Data Engineering},
  year = {2023}
  }


• Partial-Label Regression.
  X. Cheng, D.-B. Wang, L. Feng, M.-L. Zhang, B. An.
  In: Proceedings of the 37th AAAI Conference on Artificial Intelligence, Washington, DC, USA.
  
   AAAI 2023    Abstract    Paper    Appendix    Code    Bibtex  
  
  Partial-label learning is a popular weakly supervised learning setting that allows each training example to be annotated with a set of candidate labels. Previous studies on partial-label learning only focused on the classification setting where candidate labels are all discrete, which cannot handle continuous labels with real values. In this paper, we provide the first attempt to investigate partial-label regression, where each training example is annotated with a set of real-valued candidate labels. To solve this problem, we first propose a simple baseline method that takes the average loss incurred by candidate labels as the predictive loss. The drawback of this method lies in that the loss incurred by the true label may be overwhelmed by other false labels. To overcome this drawback, we propose an identification method that takes the least loss incurred by candidate labels as the predictive loss. We further improve it by proposing a progressive identification method to differentiate candidate labels using progressively updated weights for incurred losses. We prove that the latter two methods are model-consistent and provide convergence analysis showing the optimal parametric convergence rate. Our proposed methods are theoretically grounded and can be compatible with any models, optimizers, and losses. Experiments validate the effectiveness of our proposed methods. @inproceedings{AAAI23Cheng,
  author = {Xin Cheng, Deng-Bao Wang, Lei Feng, Min-Ling Zhang and Bo An},
  title = {Partial-Label Regression},
  booktitle = {Proceedings of the 37th AAAI Conference on Artificial Intelligence},
  year = {2023}
  }


• Adaptive Graph Guided Disambiguation for Partial Label Learning.
  D.-B. Wang, M.-L. Zhang, L. Li.
  IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022.
  
   IEEE TPAMI    Abstract    Paper    Appendix    Code    Bibtex  
  
  In partial label learning, a multi-class classifier is learned from the ambiguous supervision where each training example is associated with a set of candidate labels among which only one is valid. An intuitive way to deal with this problem is label disambiguation, i.e. differentiating the labeling confidences of different candidate labels so as to try to recover ground-truth labeling information. Recently, feature-aware label disambiguation has been proposed which utilizes the graph structure of feature space to generate labeling confidences over candidate labels. Nevertheless, the existence of noises and outliers in training data makes the graph structure derived from original feature space less reliable. In this paper, a novel partial label learning approach based on adaptive graph guided disambiguation is proposed, which is shown to be more effective in revealing the intrinsic manifold structure among training examples. Other than the sequential disambiguation-then-induction learning strategy, the proposed approach jointly performs adaptive graph construction, candidate label disambiguation and predictive model induction with alternating optimization. Furthermore, we consider the particular human-in-the-loop framework in which a learner is allowed to actively query some ambiguously labeled examples for manual disambiguation. Extensive experiments clearly validate the effectiveness of adaptive graph guided disambiguation for learning from partial label examples. @inproceedings{tpami22dbwang,
  author = {Deng-Bao Wang and Li Li and Min-Ling Zhang},
  title = {Adaptive Graph Guided Disambiguation for Partial Label Learning},
  booktitle = {IEEE Transactions on Pattern Analysis and Machine Intelligence, in press},
  year = {2022}
  }


• Revisiting Consistency Regularization for Deep Partial Label Learning.
  D.-D. Wu†, D.-B. Wang†, M.-L. Zhang.
  In: Proceedings of the 39th International Conference on Machine Learning, Baltimore, MD USA.
  
   ICML 2022    Abstract    Paper    Code    Bibtex  
  
  Partial label learning (PLL), which refers to the classification task where each training instance is ambiguously annotated with a set of candidate labels, has been recently studied in deep learning paradigm. Despite advances in recent deep PLL literature, existing methods (e.g., methods based on self-training or contrastive learning) are confronted with either ineffectiveness or inefficiency. In this paper, we revisit a simple idea namely consistency regularization, which has been shown effective in traditional PLL literature, to guide the training of deep models. Towards this goal, a new regularized training framework, which performs supervised learning on non-candidate labels and employs consistency regularization on candidate labels, is proposed for PLL. We instantiate the regularization term by matching the outputs of multiple augmentations of an instance to a conformal label distribution, which can be adaptively inferred by the closed-form solution. Experiments on benchmark datasets demonstrate the superiority of the proposed method compared with other state-of-the-art methods. @inproceedings{ICML22ddwu,
  author = {Dong-Dong Wu and Deng-Bao Wang and Min-Ling Zhang},
  title = {Revisiting Consistency Regularization for Deep Partial Label Learning},
  booktitle = {Proceedings of the 39th International Conference on Machine Learning},
  year = {2022}
  }


• Partial Label Learning with Emerging New Labels.
  X.-R. Yu, D.-B. Wang , M.-L. Zhang
  Machine Learning Journal, 2022.
  
   Mach. Learn. 2022    Abstract    Paper    Code    Bibtex  
  
  Partial label learning deals with the problem where each training instance is associated with a set of candidate labels, among which only one is valid. Existing approaches on partial label learning assume that the scale of label space is fixed, however, this assumption may not be satisfied in open and dynamic environment. In this paper, the first attempt towards the problem of partial label learning with emerging new labels is presented. There are mainly three challenges in this task, namely new label detection, effective classification, and efficient model updating. Specifically, a new method is proposed to address these challenges which consists of three parts: (1) An ensemble-based detector that identifies instances from new labels while also assigns candidate labels to instances which may belong to known labels. (2) An effective classification mechanism that involves data pool construction and label disambiguation process. (3) An efficient updating procedure that adapts both the detector and classifier to new labels without training from scratch. Our experiments on artificial and real-world partial label data sets validate the effectiveness of the proposed method in dealing with emerging labels for partial label learning. @inproceedings{ML22Yu,
  author = {Xiang-Ru Yu, Deng-Bao Wang and Min-Ling Zhang},
  title = {Partial Label Learning with Emerging New Labels},
  booktitle = {Machine Learning, 2022.},
  year = {2022}
  }


• Rethinking Calibration of Deep Neural Networks: Don't Be Afraid of Overconfidence.
  D.-B. Wang, L. Feng, M.-L. Zhang.
  In: Advances in Neural Information Processing Systems, Virtual Conference.
  
   NeurIPS 2021    Abstract    Paper    Appendix    Code    Bibtex  
  
  Capturing accurate uncertainty quantification of the prediction from deep neural networks is important in many real-world decision-making applications. A reliable predictor is expected to be accurate when it is confident about its predictions and indicate high uncertainty when it is likely to be inaccurate. However, modern neural networks have been found to be poorly calibrated, primarily in the direction of overconfidence. In recent years, there is a surge of research on model calibration by leveraging implicit or explicit regularization techniques during training, which obtain well calibration by avoiding overconfident outputs.
  
  In our study, we empirically found that despite the predictions obtained from these regularized models are better calibrated, they are worse calibratable, namely, it is harder to further calibrate their predictions with post-hoc calibration methods like temperature scaling and histogram binning. We conduct a series of empirical studies showing that overconfidence may not hurt final calibration performance if post-hoc calibration is allowed, rather, the penalty of confident outputs will compress the room of potential improvements in post-hoc calibration phase. Our experimental findings point out a new direction to improve calibration of DNNs by considering main training and post-hoc calibration as a unified framework. @inproceedings{neurips21dbwang,
  author = {Deng-Bao Wang and Lei Feng and Min-Ling Zhang},
  title = {Rethinking Calibration of Deep Neural Networks: Do Not Be Afraid of Overconfidence},
  booktitle = {Advances in Neural Information Processing Systems},
  year = {2021}
  }


• Learning from Complementary Labels via Partial-Output Consistency Regularization.
  D.-B. Wang, L. Feng, M.-L. Zhang.
  In: Proceedings of the 30th International Joint Conference on Artificial Intelligence, Virtual Conference.
  
   IJCAI 2021    Abstract    Paper    Code    Bibtex  
  
  In complementary-label learning (CLL), a multiclass classifier is learned from training instances each associated with complementary labels, which specify the classes that the instance does not belong to. Previous studies focus on unbiased risk estimator or surrogate loss while neglect the importance of regularization in training phase. In this  paper , we give the first attempt to leverage regularization techniques for CLL. By decoupling a label vector into complementary labels and partial unknown labels, we simultaneously inhibit the outputs of complementary labels with a complementary loss and penalize the sensitivity of the classifier on the partial outputs of these unknown classes by consistency regularization. Then we unify the complementary loss and consistency loss together by a specially designed dynamic weighting factor. We conduct a series of experiments showing that the proposed method achieves highly competitive performance in CLL. @inproceedings{ijcai21dbwang,
  author = {Deng-Bao Wang and Lei Feng and Min-Ling Zhang},
  title = {Learning from Complementary Labels via Partial-Output Consistency Regularization},
  booktitle = {Proceedings of the 30th International Joint Conference on Artificial Intelligence},
  pages = {3075--3081},
  year = {2021}
  }


• Learning from Noisy Labels with Complementary Loss Functions.
  D.-B. Wang, Y. Wen, L. Pan, M.-L. Zhang.
  In: Proceedings of the 35th AAAI Conference on Artificial Intelligence, Virtual Conference.
  
   AAAI 2021    Abstract    Paper    Appendix    Code    Bibtex  
  
  Recent researches reveal that deep neural networks are sensitive to label noises hence leading to poor generalization performance in some tasks. Although different robust loss functions have been proposed to remedy this issue, they suffer from an underfitting problem, thus are not sufficient to learn accurate models. On the other hand, the commonly used Cross Entropy (CE) loss, which shows high performance in standard supervised learning (with clean supervision), is nonrobust to label noise. In this  paper , we propose a general framework to learn robust deep neural networks with complementary loss functions. In our framework, CE and robust loss play complementary roles in a joint learning objective as per their learning sufficiency and robustness properties respectively. Specifically, we find that by exploiting the memorization effect of neural networks, we can easily filter out a proportion of hard samples and generate reliable pseudo labels for easy samples, and thus reduce the label noise to a quite low level. Then, we simply learn with CE on pseudo supervision and robust loss on original noisy supervision. In this procedure, CE can guarantee the sufficiency of optimization while the robust loss can be regarded as the supplement. Experimental results on benchmark classification datasets indicate that the proposed method helps achieve robust and sufficient deep neural network training simultaneously. @inproceedings{aaai21dbwang,
  author = {Deng-Bao Wang and Yong Wen and Lujia Pan and and Min-Ling Zhang},
  title = {Learning from Noisy Labels with Complementary Loss Functions},
  booktitle = {Proceedings of the 35th AAAI Conference on Artificial Intelligence},
  pages = {10111--10119},
  year = {2021}
  }


• Adaptive Graph Guided Disambiguation for Partial Label Learning.
  D.-B. Wang, L. Li, M.-L. Zhang.
  In: Proceedings of the 25th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Anchorage, AK, USA.
  
   KDD 2019    Abstract    Paper    Code    Bibtex  
  
  Partial label learning aims to induce a multi-class classifer from training examples where each of them is associated with a set of candidate labels, among which only one is the ground-truth label. The common strategy to train predictive model is disambiguation, i.e. differentiating the modeling outputs of individual candidate labels so as to recover ground-truth labeling information. Recently, feature-aware disambiguation was proposed to generate different labeling confdences over candidate label set by utilizing the graph structure of feature space. However, the existence of noise and outliers in training data makes the similarity derived from original features less reliable. To this end, we proposed a novel approach for partial label learning based on adaptive graph guided disambiguation (PL-AGGD). Compared with fxed graph, adaptive graph could be more robust and accurate to reveal the intrinsic manifold structure within the data. Moreover, instead of the two-stage strategy in previous algorithms, our approach performs label disambiguation and predictive model training simultaneously. Specifically, we present a unifed framework which jointly optimizes the ground-truth labeling confdences, similarity graph and model parameters to achieve strong generalization performance. Extensive experiments show that PL-AGGD performs favorably against stateof-the-art partial label learning approaches. @inproceedings{kdd19dbwang,
  author = {Deng-Bao Wang and Li Li and Min-Ling Zhang},
  title = {Adaptive Graph Guided Disambiguation for Partial Label Learning},
  booktitle = {Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Anchorage, AK, USA},
  pages = {83--91},
  year = {2019}
  }


• Multi-View Multi-Label Learning with View-Specific Information Extraction.
  X. Wu, Q.-G. Chen, Y. Hu, D.-B. Wang, X. Chang, X. Wang, M.-L. Zhang.
  In: Proceedings of the 28th International Joint Conference on Artificial Intelligence, Macao, China.
  
   IJCAI 2019    Abstract    Paper    Bibtex  
  
  Multi-view multi-label learning serves an important framework to learn from objects with diverse representations and rich semantics. Existing multi-view multi-label learning techniques focus on exploiting shared subspace for fusing multiview representations, where helpful view-specific information for discriminative modeling is usually ignored. In this  paper , a novel multi-view multi-label learning approach named SIMM is proposed which leverages shared subspace exploitation and view-specific information extraction. For shared subspace exploitation, SIMM jointly minimizes confusion adversarial loss and multi-label loss to utilize shared information from all views. For view-specific information extraction, SIMM enforces an orthogonal constraint w.r.t. the shared subspace to utilize view-specific discriminative information. Extensive experiments on real-world data sets clearly show the favorable performance of SIMM against other state-of-the-art multi-view multi-label learning approaches @inproceedings{ijcai19jhwu,
  author = {Xuan Wu and Qing-Guo Chen and Yao Hu and Deng-Bao Wang and Xiaodong Chang and Xiaobo Wang and Min-Ling Zhang},
  title = {Multi-View Multi-Label Learning with View-Specific Information Extraction},
  booktitle = {Proceedings of the 28th International Joint Conference on Artificial Intelligence, Macao, China},
  pages = {3884--3890},
  year = {2019}
  }

Honors

• DAAD AInet Fellow   2023

• National Scholarship   2022

• Merit Student Award, Southeast University   2022

• Tencent Rhino-Bird Elite Training Program  2022

• Special Freshman Scholarship for PhD Students   2019

• Outstanding Graduates, Southwest University   2019

• National Scholarship   2018

• Merit Student Award, Southwest University   2018

• First Class Academic Scholarship, Southwest University   2016, 2017, 2018

Academic Services

• Reviewer for ICLR (2024) ICML (2022, 2023), IJCAI (2022, 2023), AAAI (2021, 2022, 2024), ECML/PKDD (2022), ACML (2021), ICMLA (2021), IAAI (2021, 2022).

• Invited Journal Reviewer for IEEE TPAMI, SCIENCE CHINA Information Sciences, ACM TIST, ACM TKDD, IEEE TMM, JCST, Neurocomputing.