arxiv: 2604.03306 · v1 · submitted 2026-03-31 · 💻 cs.CV

Recognition: 2 theorem links

· Lean Theorem

Deep Image Clustering Based on Curriculum Learning and Density Information

Haiyang Zheng , Ruilin Zhang , Hongpeng Wang

Authors on Pith no claims yet

Pith reviewed 2026-05-13 23:59 UTC · model grok-4.3

classification 💻 cs.CV

keywords image clusteringdeep clusteringcurriculum learningdensity informationdensity coreIDCLrobust clusteringcluster assignment

0 comments

The pith

A deep clustering method trains on density-ordered examples first and assigns points via density cores rather than cluster centers.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces IDCL, a deep image clustering approach that adds a curriculum learning schedule driven by density information to set the order and pace of model updates. It also replaces direct distance-to-center assignment with guidance from density cores to limit error buildup during iterations. If the claims hold, the method yields more robust partitions on complex image data, faster convergence, and greater tolerance to changes in dataset size or cluster count. Existing deep clustering methods skip explicit learning strategies and accumulate errors from repeated point-to-center decisions. The authors test the resulting pipeline on standard benchmarks and report gains in accuracy and stability.

Core claim

The central claim is that a curriculum learning scheme grounded in input density information supplies a more reasonable training pace, and that substituting density cores for individual cluster centers to guide assignment reduces error accumulation across iterations, producing a clustering method (IDCL) that is more robust than prior deep approaches on image data.

What carries the argument

Curriculum learning schedule ordered by density information, together with density-core guidance that replaces point-to-center distance for cluster assignment.

If this is right

The method converges faster because early training focuses on high-density examples that are easier to separate.
Error accumulation drops when assignments follow density cores instead of single centers, yielding more stable cluster labels over iterations.
The approach adapts to varying numbers of clusters and data scales without changing the core density-driven schedule.
Robustness improves across different image contexts because the curriculum and core mechanisms do not rely on point-wise distances alone.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same density-ordered curriculum could be tested on non-image data such as text embeddings or time-series to check whether the ordering benefit generalizes.
Density cores might serve as a drop-in replacement for centers in other iterative clustering algorithms that currently use k-means style assignments.
If density estimation itself is noisy on very high-dimensional inputs, the curriculum schedule could be combined with a preliminary dimensionality reduction step.

Load-bearing premise

Density values computed from the data give a reliably better order and pace for training than standard schedules, and switching to density cores measurably cuts error without creating new instabilities.

What would settle it

Run IDCL and a baseline deep clustering method on the same benchmark images; if the density-based curriculum and core assignment produce no measurable rise in final clustering accuracy or no drop in required epochs to converge, the central claim is false.

Figures

Figures reproduced from arXiv: 2604.03306 by Haiyang Zheng, Hongpeng Wang, Ruilin Zhang.

**Figure 1.** Figure 1: The framework of our proposed IDCL. For𝑇 > 𝑡, we find that𝜉2 > 𝜉1. As 𝐹 ′′ 𝜁 (𝑡) > 0, this implies that 𝐹 ′ 𝜁 (𝑡) is monotonically increasing, and thus, 𝐹 ′ 𝜁 (𝜉2) > 𝐹 ′ 𝜁 (𝜉1). Therefore, if 𝜖2 > 0, we obtain: (𝐹𝜁 (𝑇 + 𝜖2) − 𝐹𝜁 (𝑇 )) > (𝐹𝜁 (𝑡 + 𝜖2) − 𝐹𝜁 (𝑡)). Consequently, for 𝜁 𝑡 = 𝐹𝜁 (𝑡), we have: 𝑛 · 𝜁 𝑡+𝜖1 ≥ 𝑛 · 𝜁 𝑡 , s.t. 𝜖1 > 0; (𝑛 · 𝜁 𝑇+𝜖2 − 𝑛 · 𝜁 𝑇 ) ≥ (𝑛 · 𝜁 𝑡+𝜖2 − 𝑛 · 𝜁 𝑡 ), s.t. 𝑇 > 𝑡, 𝜖2 > 0 B… view at source ↗

**Figure 2.** Figure 2: Schematic illustration of Cluster Assignment on [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Figure 3: The convergence process on MNIST-test dataset. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: Output results of the Decoder on YTF dataset. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

**Figure 5.** Figure 5: Datasets visualization [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗

read the original abstract

Image clustering is one of the crucial techniques in multimedia analytics and knowledge discovery. Recently, the Deep clustering method (DC), characterized by its ability to perform feature learning and cluster assignment jointly, surpasses the performance of traditional ones on image data. However, existing methods rarely consider the role of model learning strategies in improving the robustness and performance of clustering complex image data. Furthermore, most approaches rely solely on point-to-point distances to cluster centers for partitioning the latent representations, resulting in error accumulation throughout the iterative process. In this paper, we propose a robust image clustering method (IDCL) which, to our knowledge for the first time, introduces a model training strategy using density information into image clustering. Specifically, we design a curriculum learning scheme grounded in the density information of input data, with a more reasonable learning pace. Moreover, we employ the density core rather than the individual cluster center to guide the cluster assignment. Finally, extensive comparisons with state-of-the-art clustering approaches on benchmark datasets demonstrate the superiority of the proposed method, including robustness, rapid convergence, and flexibility in terms of data scale, number of clusters, and image context.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper adds density-based curriculum ordering and density-core assignments to deep image clustering to reduce error buildup, but the abstract shows no numbers so the gains stay unproven.

read the letter

The main thing to know is that this paper introduces density information into the training process for deep image clustering. They build a curriculum that orders samples by density so the model starts with clearer examples, and they replace standard cluster centers with density cores when assigning points. The goal is to slow early error accumulation and improve stability across different data scales and cluster counts. That combination is flagged as new for this setting, and the logic lines up with known issues in iterative deep clustering methods where bad early assignments compound. The abstract does a clean job of naming the problems with point-to-point center distances and proposing a direct fix. If the experiments hold, the robustness claims would be useful for multimedia applications where image data varies a lot. The soft spots are straightforward. The summary gives no quantitative results, no ablation tables, and no details on how density is estimated in the latent space or how the curriculum schedule is implemented. Without those, it is impossible to tell whether the reported superiority comes from the density ideas or from other tuning choices. Density estimation itself can be noisy in high-dimensional features, so any instability there would need checking. The full paper likely contains the benchmark numbers and code, but on the given text the central claims rest on unshown evidence. This work is for people already working on unsupervised image clustering who want incremental training tweaks rather than new architectures. A reader focused on practical robustness in DC methods could pick up usable ideas even if the exact gains need verification. I would send it for peer review. The pipeline is coherent, the addressed problems are real, and referees can evaluate the results and reproducibility directly.

Referee Report

1 major / 2 minor

Summary. The manuscript proposes IDCL, a deep image clustering approach that introduces density information to define a curriculum learning schedule for training pace and replaces per-sample distances to cluster centers with density-core guidance for assignment. The method is claimed to reduce iterative error accumulation while improving robustness to scale, cluster count, and image context, with extensive benchmark comparisons demonstrating superiority over prior deep clustering methods.

Significance. If the reported gains hold under scrutiny, the work offers a plausible new training strategy for deep clustering by grounding curriculum order and assignment in density estimates rather than raw distances. This directly targets the error-accumulation problem noted in the abstract and could be useful for complex image data where standard schedules and center-based losses are brittle.

major comments (1)

[§4] §4 (Experiments) and associated tables: the abstract asserts benchmark superiority, reduced error accumulation, and robustness, yet the provided text supplies no quantitative metrics, error bars, ablation results isolating the density curriculum and density-core components, or statistical significance tests. Without these, the central empirical claim cannot be evaluated and the attribution of gains remains unsupported.

minor comments (2)

[§3] Notation for the density estimator and curriculum weighting function is introduced without an explicit equation or pseudocode; adding these would clarify how density is computed and used to order samples.
[§2] The claim of being 'to our knowledge for the first time' to introduce density-based curriculum into image clustering should be supported by a brief related-work comparison table or explicit citation contrast.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed review and constructive feedback. We address the single major comment below and will incorporate the suggested improvements into the revised manuscript.

read point-by-point responses

Referee: [§4] §4 (Experiments) and associated tables: the abstract asserts benchmark superiority, reduced error accumulation, and robustness, yet the provided text supplies no quantitative metrics, error bars, ablation results isolating the density curriculum and density-core components, or statistical significance tests. Without these, the central empirical claim cannot be evaluated and the attribution of gains remains unsupported.

Authors: We agree that the current manuscript version lacks error bars, component-wise ablations, and statistical significance tests, which limits the strength of the empirical claims. In the revision we will add: (1) mean and standard deviation over five independent runs for all reported metrics (ACC, NMI, ARI) on the benchmark tables; (2) a dedicated ablation table that isolates the contribution of the density-based curriculum schedule and the density-core guidance mechanism; and (3) paired t-test p-values comparing IDCL against the strongest baselines to establish statistical significance. These additions will directly support the claims of reduced error accumulation and improved robustness. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The provided abstract and description present IDCL as a new pipeline that applies density-based curriculum ordering and density-core assignment in place of standard center distances. No equations, fitted parameters, or self-citations are shown that would reduce the claimed improvements to quantities defined by the same data or prior author work. The method is framed as importing external density concepts into clustering, with benefits asserted via empirical comparison rather than internal redefinition. This keeps the central claims independent of the inputs they operate on.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities; the method appears to rest on standard deep-learning assumptions and density concepts drawn from prior literature.

pith-pipeline@v0.9.0 · 5494 in / 958 out tokens · 35542 ms · 2026-05-13T23:59:19.517950+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

we design a curriculum learning scheme grounded in the density information of input data... employ the density core rather than the individual cluster center
IndisputableMonolith/Foundation/AbsoluteFloorClosure.lean absolute_floor_iff_bare_distinguishability unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

ρ_i = Σ exp(−‖z_i−z_j‖² / d_c²)

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

92 extracted references · 92 canonical work pages · 2 internal anchors

[1]

Deep embedded median clustering for routing misbehaviour and attacks detection in ad-hoc networks

2022. Deep embedded median clustering for routing misbehaviour and attacks detection in ad-hoc networks. Ad Hoc Networks 126 (2022), 102757

work page 2022
[2]

Yoshua Bengio, Jérôme Louradour, Ronan Collobert, and Jason Weston

work page
[3]

ISBN 9781605585161

Curriculum Learning. In Proceedings of the 26th Annual International Conference on Machine Learning (Montreal, Quebec, Canada) (ICML ’09). As- sociation for Computing Machinery, New York, NY, USA, 41¨C48. https://doi. org/10.1145/1553374.1553380

work page doi:10.1145/1553374.1553380
[4]

Jinyu Cai, Jicong Fan, Wenzhong Guo, Shiping Wang, Yunhe Zhang, and Zhao Zhang. 2022. Efficient Deep Embedded Subspace Clustering. In Proceedings of the IEEE/CVF Conferenceon Computer Vision and Pattern Recognition. 1–10

work page 2022
[5]

Mathilde Caron, Piotr Bojanowski, Armand Joulin, and Matthijs Douze. 2018. Deep clustering for unsupervised learning of visual features. In Proceedings of the European conferenceon computer vision (ECCV). 132–149

work page 2018
[7]

Rui Chen, Yongqiang Tang, Lei Tian, Caixia Zhang, and Wensheng Zhang. 2022. Deep convolutional self-paced clustering. Applied Intelligence (2022), 1–15

work page 2022
[8]

Adam Coates, Andrew Ng, and Honglak Lee. 2011. An Analysis of Single- Layer Networks in Unsupervised Feature Learning. In Proceedings of the Fourteenth International Conference on Artificial Intelligence and Statistics (Proceedings of Machine Learning Research, Vol. 15), Geoffrey Gordon, David Dunson, and Miroslav Dud¨ªk (Eds.). PMLR, Fort Lauderdale, FL,...

work page 2011
[9]

Gregory Cohen, Saeed Afshar, Jonathan Tapson, and Andre Van Schaik. 2017. EMNIST: Extending MNIST to handwritten letters. In 2017 international joint conference on neural networks (IJCNN). IEEE, 2921–2926. https://doi.org/10. 1109/IJCNN.2017.7966217

work page arXiv 2017
[10]

Yao Ding, Zhili Zhang, Xiaofeng Zhao, Wei Cai, Nengjun Yang, Haojie Hu, Xi- anxiang Huang, Yuan Cao, and Weiwei Cai. 2022. Unsupervised Self-Correlated Learning Smoothy Enhanced Locality Preserving Graph Convolution Embed- ding Clustering for Hyperspectral Images.IEEE Transactions on Geoscience and Remote Sensing 60 (2022), 1–16. https://doi.org/10.1109/T...

work page doi:10.1109/tgrs.2022.3202865 2022
[11]

Dheeru Dua and Casey Graff. 2017. UCI Machine Learning Repository. http: //archive.ics.uci.edu/ml

work page 2017
[12]

Liang Duan, Charu Aggarwal, Shuai Ma, and Saket Sathe. 2019. Improving Spectral Clustering with Deep Embedding and Cluster Estimation. In 2019 IEEE International Conferenceon Data Mining (ICDM). 170–179. https://doi.org/10. 1109/ICDM.2019.00027

work page arXiv 2019
[13]

Johan Edstedt, Georg Bökman, and Zhenjun Zhao. 2024. Dedode v2: Analyzing and improving the dedode keypoint detector. In Proceedings of the IEEE/CVF Conferenceon Computer Vision and Pattern Recognition. 4245–4253

work page 2024
[14]

Matheus Campos Fernandes, Thiago Ferreira Cov?es, and Andr¨¦ Luiz Vizine Pereira. 2020. Improving evolutionary constrained clustering using Active Learning. Knowledge-Based Systems 209 (2020), 106452

work page 2020
[15]

Kamran Ghasedi, Xiaoqian Wang, Cheng Deng, and Heng Huang. 2019. Bal- anced Self-Paced Learning for Generative Adversarial Clustering Network. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

work page 2019
[16]

Xifeng Guo, Long Gao, Xinwang Liu, and Jianping Yin. 2017. Improved Deep Embedded Clustering with Local Structure Preservation. In Proceedings ofthe 26th International Joint Conferenceon Artificial Intelligence (Melbourne, Aus- tralia) (IJCAI’17). AAAI Press, 1753¨C1759

work page 2017
[17]

Xifeng Guo, Xinwang Liu, En Zhu, Xinzhong Zhu, Miaomiao Li, Xin Xu, and Jian- ping Yin. 2020. Adaptive Self-Paced Deep Clustering with Data Augmentation. IEEE Transactions on Knowledge and Data Engineering 32, 9 (2020), 1680–1693. https://doi.org/10.1109/TKDE.2019.2911833

work page doi:10.1109/tkde.2019.2911833 2020
[18]

Xifeng Guo, En Zhu, Xinwang Liu, and Jianping Yin. 2018. Deep Embedded Clus- tering with Data Augmentation. In Proceedings ofThe 10th Asian Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 95), Jun Zhu and Ichiro Takeuchi (Eds.). PMLR, 550–565

work page 2018
[19]

Kai Han, Andrea Vedaldi, and Andrew Zisserman. 2019. Learning to Dis- cover Novel Visual Categories via Deep Transfer Clustering. In 2019 IEEE/CVF International Conferenceon Computer Vision (ICCV). 8400–8408. https://doi. org/10.1109/ICCV.2019.00849

work page doi:10.1109/iccv.2019.00849 2019
[20]

Sebastian Houben, Johannes Stallkamp, Jan Salmen, Marc Schlipsing, and Chris- tian Igel. 2013. Detection of traffic signs in real-world images: The German Traf- fic Sign Detection Benchmark. In The 2013 international joint conference on neural networks (IJCNN). 1–8. https://doi.org/10.1109/IJCNN.2013.6706807

work page doi:10.1109/ijcnn.2013.6706807 2013
[21]

Jonathan J. Hull. 1994. A database for handwritten text recognition research. IEEE Transactionson pattern analysis and machine intelligence 16, 5 (1994), 550–

work page 1994
[22]

https://doi.org/10.1109/34.291440

work page doi:10.1109/34.291440
[23]

Wenhao Jiang, Wei Liu, and Fu lai Chung. 2018. Knowledge transfer for spectral clustering. Pattern Recognition 81 (2018), 484–496. https://www.sciencedirect. com/science/article/pii/S0031320318301511

work page 2018
[24]

Yangbangyan Jiang, Zhiyong Yang, Qianqian Xu, Xiaochun Cao, and Qingming Huang. 2018. When to Learn What: Deep Cognitive Subspace Clustering. In Proceedings of the 26th ACM International Conferenceon Multimedia (Seoul, Republic of Korea) (MM ’18). Association for Computing Machinery, New York, NY, USA, 718¨C726. https://doi.org/10.1145/3240508.3240582

work page doi:10.1145/3240508.3240582 2018
[25]

Zhuxi Jiang, Yin Zheng, Huachun Tan, Bangsheng Tang, and Hanning Zhou

work page
[26]

In Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence, IJCAI-17

Variational Deep Embedding: An Unsupervised and Generative Ap- proach to Clustering. In Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence, IJCAI-17. 1965–1972. https://doi.org/10. 24963/ijcai.2017/273

work page 1965
[27]

Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic opti- mization. arXiv preprint arXiv:1412.6980 (2014)

work page internal anchor Pith review Pith/arXiv arXiv 2014
[28]

Qizhen Lan and Qing Tian. 2024. Gradient-guided knowledge distillation for object detectors. In Proceedings of the IEEE/CVF Winter Conference on Applications ofComputer Vision. 424–433

work page 2024
[29]

Qizhen Lan and Qing Tian. 2025. ACAM-KD: adaptive and cooperative at- tention masking for knowledge distillation. In Proceedings of the IEEE/CVF International Conferenceon Computer Vision. 3957–3966

work page 2025
[30]

Qizhen Lan and Qing Tian. 2026. CLoCKDistill: Consistent Location and Con- text aware Knowledge Distillation for DETRs. In Proceedings ofthe IEEE/CVF Winter Conferenceon Applications ofComputer Vision. 7188–7197

work page 2026
[31]

Yann LeCun, Léon Bottou, Yoshua Bengio, and Patrick Haffner. 1998. Gradient- based learning applied to document recognition. Proc. IEEE 86, 11 (1998), 2278–

work page 1998
[32]

https://doi.org/10.1109/5.726791

work page doi:10.1109/5.726791
[33]

Collin Leiber, Lena GM Bauer, Benjamin Schelling, Christian Böhm, and Clau- dia Plant. 2021. Dip-based deep embedded clustering with k-estimation. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining (KDD ’21, 11). Association for Computing Machinery, New York, NY, USA, 903–913. https://doi.org/10.1145/3447548.3467316

work page doi:10.1145/3447548.3467316 2021
[34]

David D Lewis, Yiming Yang, Tony Russell-Rose, and Fan Li. 2004. Rcv1: A new benchmark collection for text categorization research. Journal ofmachine learning research 5, Apr (2004), 361–397

work page 2004
[35]

Bingyu Li, Haocheng Dong, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xuelong Li. 2025. Exploring efficient open-vocabulary segmentation in the remote sens- ing. arXiv preprint arXiv:2509.12040 (2025)

work page arXiv 2025
[36]

Bingyu Li, Tao Huo, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xuelong Li. 2025. Exploring the Underwater World Segmentation without Extra Training. arXiv preprint arXiv:2511.07923 (2025)

work page arXiv 2025
[37]

Bingyu Li, Feiyu Wang, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xuelong Li

work page
[38]

Maris: Marine open-vocabulary in- stance segmentation with geometric enhancement and se- mantic alignment.arXiv preprint arXiv:2510.15398, 2025

MARIS: Marine Open-Vocabulary Instance Segmentation with Geometric Enhancement and Semantic Alignment. arXiv preprint arXiv:2510.15398 (2025)

work page arXiv 2025
[39]

Bingyu Li, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xuelong Li. 2025. Stitchfu- sion: Weaving any visual modalities to enhance multimodal semantic segmenta- tion. In Proceedings ofthe 33rd ACM International Conferenceon Multimedia. 1308–1317

work page 2025
[40]

Hongyu Li, Lefei Zhang, and Kehua Su. 2023. Dual Mutual Information Con- straints for Discriminative Clustering. Proceedings ofthe AAAI Conferenceon ArtificialIntelligence 37, 7 (Jun. 2023), 8571–8579. https://doi.org/10.1609/aaai. v37i7.26032

work page doi:10.1609/aaai 2023
[41]

Haoang Li, Ji Zhao, Jean-Charles Bazin, Pyojin Kim, Kyungdon Joo, Zhenjun Zhao, and Yun-Hui Liu. 2023. Hong kong world: Leveraging structural regu- larity for line-based slam. IEEE Transactions on Pattern Analysis and Machine Intelligence 45, 11 (2023), 13035–13053

work page 2023
[42]

Muquan Li, Qian Dong, Dongyang Zhang, Ke Qin, and Guangchun Luo. 2026. Efficient Industrial Dataset Distillation With Textual Trajectory Matching.IEEE Transactions on Industrial Informatics (2026)

work page 2026
[43]

Muquan Li, Hang Gou, Yingyi Ma, Rongzheng Wang, Ke Qin, and Tao He. 2026. Fixed Anchors Are Not Enough: Dynamic Retrieval and Persistent Homology for Dataset Distillation. arXiv preprint arXiv:2602.24144 (2026)

work page arXiv 2026
[44]

Muquan Li, Hang Gou, Dongyang Zhang, Shuang Liang, Xiurui Xie, Deqiang Ouyang, and Ke Qin. 2025. Beyond Random: Automatic Inner-loop Optimization ICMR ’24, June 10–14, 2024, Phuket, Thailand Haiyang Zheng, Ruilin Zhang, & Hongpeng Wang in Dataset Distillation. In NeurIPS

work page 2025
[45]

M. Li, D. Zhang, Q. Dong, et al. 2025. Adaptive Dataset Quantization. In Proceedings ofthe AAAI Conferenceon Artificial Intelligence, Vol. 39. 12093– 12101

work page 2025
[46]

Muquan Li, Dongyang Zhang, Tao He, Xiurui Xie, Yuan-Fang Li, and Ke Qin

work page
[47]

In Proceedings of the 32nd ACM International Conference on Multimedia, MM

Towards Effective Data-Free Knowledge Distillation via Diverse Diffusion Augmentation. In Proceedings of the 32nd ACM International Conference on Multimedia, MM. 4416–4425

work page
[48]

Sicheng Li, Yuhui Chu, Yunpeng Zhao, and Pengpeng Zhao. 2024. An efficient deep learning-based framework for image distortion correction. The Visual Computer 40, 10 (2024), 6955–6967

work page 2024
[49]

Sicheng Li, Mai Dan, Yuhui Chu, Jiahui Yu, Yunpeng Zhao, and Pengpeng Zhao

work page
[50]

In International Conference on Medical Image Computing and Computer-Assisted Intervention

RetiDiff: Diffusion-Based Synthesis of Retinal OCT Images for Enhanced Segmentation. In International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 516–525

work page
[51]

Sicheng Li, Mai Dan, Yuhui Chu, Yunpeng Zhao, and Pengpeng Zhao. 2026. Cross-domain knowledge integration framework for optical coherence tomog- raphy images denoising. Biomedical Signal Processing and Control 113 (2026), 108916

work page 2026
[52]

Sicheng Li, Yunpeng Zhao, and Pengpeng Zhao. 2026. GarNet: Geometry- Aware Rectification Network for Generic Image Distortions. Knowledge-Based Systems (2026), 115632

work page 2026
[53]

Wei Li, Bing Hu, Rui Shao, Leyang Shen, and Liqiang Nie. 2025. Lion-fs: Fast & slow video-language thinker as online video assistant. In Proceedings ofthe IEEE/CVF Conferenceon Computer Vision and Pattern Recognition. 3240–3251

work page 2025
[54]

Wei Li, Renshan Zhang, Rui Shao, Jie He, and Liqiang Nie. 2025. Cogvla: Cognition-aligned vision-language-action model via instruction-driven routing & sparsification. In Advances in Neural InformationProcessing Systems

work page 2025
[55]

Yunfan Li, Peng Hu, Zitao Liu, Dezhong Peng, Joey Tianyi Zhou, and Xi Peng

work page
[56]

Proceedings ofthe AAAI Conferenceon Artificial Intelligence 35, 10 (May 2021), 8547–8555

Contrastive Clustering. Proceedings ofthe AAAI Conferenceon Artificial Intelligence 35, 10 (May 2021), 8547–8555. https://doi.org/10.1609/aaai.v35i10. 17037

work page doi:10.1609/aaai.v35i10 2021
[57]

Bangyan Liao, Zhenjun Zhao, Lu Chen, Haoang Li, Daniel Cremers, and Pei- dong Liu. 2024. GlobalPointer: Large-Scale Plane Adjustment with Bi-Convex Relaxation. In European Conferenceon Computer Vision. Springer, 360–376

work page 2024
[58]

Dingyuan Liu, Qiannan Shen, and Jiaci Liu. 2026. The Health-Wealth Gradient in Labor Markets: Integrating Health, Insurance, and Social Metrics to Predict Employment Density. Computation 14, 1 (2026), 22

work page 2026
[59]

Hu Lu, Chao Chen, Hui Wei, Zhongchen Ma, Ke Jiang, and Yingquan Wang

work page
[60]

Matthew Weinberg

Improved deep convolutional embedded clustering with re-selectable sam- ple training. Pattern Recognition 127 (2022), 108611. https://doi.org/10.1016/j. patcog.2022.108611

work page doi:10.1016/j 2022
[61]

Xin Ma and Won Hwa Kim. 2022. Locally Normalized Soft Contrastive Clus- tering for Compact Clusters. In Proceedings of the Thirty-First International Joint Conferenceon Artificial Intelligence, IJCAI-22, Lud De Raedt (Ed.). Inter- national Joint Conferences on Artificial Intelligence Organization, 3313–3320. https://doi.org/10.24963/ijcai.2022/460 Main Track

work page doi:10.24963/ijcai.2022/460 2022
[62]

Daniel P. M. de Mello, Renato M. Assun⁇o, and Fabricio Murai. 2022. Top-Down Deep Clustering with Multi-Generator GANs. IEEE Transactions on Neural Networks and Learning Systems 36 (Jun 2022), 7770–7778. https://doi.org/10. 1609/aaai.v36i7.20745

work page 2022
[63]

Pavlidis

Hankui Peng and Nicos G. Pavlidis. 2019. Subspace Clustering with Active Learn- ing. In 2019 IEEE International Conferenceon Big Data (Big Data). 135–144

work page 2019
[64]

Wentao Qu, Guofeng Mei, Jing Wang, Yujiao Wu, Xiaoshui Huang, and Liang Xiao. 2025. Robust Single-Stage Fully Sparse 3D Object Detection via Detachable Latent Diffusion.arXiv preprint arXiv:2508.03252 (2025)

work page arXiv 2025
[65]

Wentao Qu, Yuantian Shao, Lingwu Meng, Xiaoshui Huang, and Liang Xiao

work page
[66]

In Proceedings ofthe IEEE/CVF Conferenceon Computer Vision and Pattern Recognition

A conditional denoising diffusion probabilistic model for point cloud up- sampling. In Proceedings ofthe IEEE/CVF Conferenceon Computer Vision and Pattern Recognition. 20786–20795

work page
[67]

Wentao Qu, Jing Wang, YongShun Gong, Xiaoshui Huang, and Liang Xiao. 2025. An end-to-end robust point cloud semantic segmentation network with single- step conditional diffusion models. In Proceedings ofthe Computer Vision and Pattern Recognition Conference. 27325–27335

work page 2025
[68]

Alex Rodriguez and Alessandro Laio. 2014. Clustering by fast search and find of density peaks. Science 344, 6191 (2014), 1492–1496. https://doi.org/10.1126/ science.1242072

work page 2014
[69]

Meitar Ronen, Shahaf E Finder, and Oren Freifeld. 2022. DeepDPM: Deep Clus- tering With an Unknown Number of Clusters. In Proceedings ofthe IEEE/CVF Conferenceon Computer Vision and Pattern Recognition. 9861–9870

work page 2022
[70]

Zhendong Ruan, Teng Yan, Yaobang Cai, Yu Han, Leyi Zheng, and Yang Zhang. 2024. Q-value regularized decision convformer for offline reinforcement learning. In 2024 IEEE International Conferenceon Robotics and Biomimetics (ROBIO). IEEE, 91–97

work page 2024
[71]

Mohammadreza Sadeghi and Narges Armanfard. 2021. IDECF: Improved Deep Embedding Clustering With Deep Fuzzy Supervision. In2021 IEEE International Conference on Image Processing (ICIP). 1009–1013. https://doi.org/10.1109/ ICIP42928.2021.9506051

work page arXiv 2021
[72]

Qiannan Shen and Jing Zhang. 2025. AI-Enhanced Disaster Risk Prediction with Explainable SHAP Analysis: A Multi-Class Classification Approach Using XG- Boost. (December 2025). https://doi.org/10.21203/rs.3.rs-8437180/v1 Preprint, Version 1, posted December 31, 2025

work page doi:10.21203/rs.3.rs-8437180/v1 2025
[73]

Qiannan Shen and Jing Zhang. 2026. MFTFormer: Meteorological-Frequency- Temporal Transformer with Block-Aligned Fusion for Traffic Flow Prediction. Research Square (2026). Preprint, doi:10.21203/rs.3.rs-8770196/v1

work page doi:10.21203/rs.3.rs-8770196/v1 2026
[74]

Haoming Song, Delin Qu, Yuanqi Yao, Qizhi Chen, Qi Lv, Yiwen Tang, Modi Shi, Guanghui Ren, Maoqing Yao, Bin Zhao, et al. 2025. Hume: Introducing System- 2 Thinking in Visual-Language-Action Model. arXiv preprint arXiv:2505.21432 (2025)

work page arXiv 2025
[75]

Aaron Xuxiang Tian, Ruofan Zhang, Janet Tang, Ji Wang, Tianyu Shi, and Jiaxin Wen. 2025. Measuring Harmfulness of Computer-Using Agents. arXiv preprint arXiv:2508.00935 (2025)

work page arXiv 2025
[76]

Tian Tian, Jie Zhang, Xiang Lin, Zhi Wei, and Hakon Hakonarson. 2021. Model- based deep embedding for constrained clustering analysis of single cell RNA-seq data. Nature communications 12, 1 (2021), 1873

work page 2021
[77]

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Ł ukasz Kaiser, and Illia Polosukhin. 2017. Attention is All you Need. In Advances in Neural Information Processing Systems, I. Guyon, U. Von Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Gar- nett (Eds.), Vol. 30. Curran Associates, Inc.https://pr...

work page 2017
[78]

Xin Wang, Yudong Chen, and Wenwu Zhu. 2022. A Survey on Curriculum Learn- ing. IEEE Transactionson Pattern Analysis and Machine Intelligence 44, 9 (2022), 4555–4576. https://doi.org/10.1109/TPAMI.2021.3069908

work page doi:10.1109/tpami.2021.3069908 2022
[79]

Xuqin Wang, Tao Wu, Yanfeng Zhang, Lu Liu, Mingwei Sun, Yongliang Wang, Niclas Zeller, and Daniel Cremers. 2026. GeodesicNVS: Probability Density Geo- desic Flow Matching for Novel View Synthesis.arXiv preprint arXiv:2603.01010 (2026)

work page arXiv 2026
[80]

Xuqin Wang, Tao Wu, Yanfeng Zhang, Lu Liu, Dong Wang, Mingwei Sun, Yongliang Wang, Niclas Zeller, and Daniel Cremers. 2025. LADB: Latent Aligned Diffusion Bridges for Semi-Supervised Domain Translation. InDAGM German Conferenceon Pattern Recognition. Springer, 221–236

work page 2025
[81]

Lior Wolf, Tal Hassner, and Itay Maoz. 2011. Face recognition in unconstrained videos with matched background similarity. InCVPR 2011. IEEE, 529–534. https: //doi.org/10.1109/CVPR.2011.5995566

work page doi:10.1109/cvpr.2011.5995566 2011

Showing first 80 references.