Efficient Federated Incomplete Multi-View Clustering

Authors: Suyuan Liu, Hao Yu, Hao Tan, Ke Liang, Siwei Wang, Shengju Yu, En Zhu, Xinwang Liu

ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on benchmark datasets demonstrate the superiority of EFIMVC in clustering accuracy, communication efficiency, and privacy preservation. Our code is publicly available at https://github.com/Tracesource/EFIMVC. ... We conduct experiments on seven multi-view datasets: Protein Fold, Web KB, 100Leaves, CCV, YTF10, CIFAR10, and MNIST, with detailed descriptions provided in Table 1.
Researcher Affiliation Academia 1College of Computer Science and Technology, National University of Defence Technology, Changsha, China 2Academy of Military Sciences, Beijing, China. Correspondence to: Xinwang Liu <EMAIL>.
Pseudocode Yes Algorithm 1 The proposed EFIMVC
Open Source Code Yes Our code is publicly available at https://github.com/Tracesource/EFIMVC.
Open Datasets Yes We conduct experiments on seven multi-view datasets: Protein Fold, Web KB, 100Leaves, CCV, YTF10, CIFAR10, and MNIST, with detailed descriptions provided in Table 1. ... 1http://mkl.ucsd.edu/dataset/protein-fold-prediction/ 2http://www.cs.umd.edu/sen/lbc-proj/LBC.html 3https://www.archive.ics.uci.edu/dataset/241 4https://www.ee.columbia.edu/ln/dvmm/CCV/ 5https://www.micc.unifi.it/resources/datasets/e-ytf/ 6http://www.cs.toronto.edu/kriz/cifar.html 7http://yann.lecun.com/exdb/mnist/
Dataset Splits No Following the definition in (Wang et al., 2022), we generate nine versions of each dataset with missing rates increasing in 10% increments, ensuring that no sample is entirely missing from all views. ... Additionally, for methods that obtain final results through k-means, we repeat the clustering process 20 times and report the average performance to mitigate the impact of initialization randomness.
Hardware Specification Yes All experiments are conducted on a computer with Intel Core i9-10900X CPU and 64G RAM.
Software Dependencies No Eq. (10) is a standard quadratic programming problem, which we solve using existing software packages. ... Similarly, we solve Eq. (14) with the quadratic programming package.
Experiment Setup Yes EFIMVC has two hyperparameters, λ and β, with search ranges of [0.001, 0.1, 1, 100]. ... The parameter settings in EFIMVC are grid search. For both λ and β, we search them in [0.001, 0.1, 1, 100]. ... for federated multi-view clustering algorithms that cannot handle missing views, we fill the missing entries with zeros before inputting the data. Additionally, for methods that obtain final results through k-means, we repeat the clustering process 20 times and report the average performance to mitigate the impact of initialization randomness.