One Pass Late Fusion Multi-view Clustering
Authors: Xinwang Liu, Li Liu, Qing Liao, Siwei Wang, Yi Zhang, Wenxuan Tu, Chang Tang, Jiyuan Liu, En Zhu
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Comprehensive experiments on multiple benchmark datasets demonstrate the superiority of our algorithm in terms of both clustering accuracy and computational efficiency. |
| Researcher Affiliation | Academia | 1School of Computer, National University of Defense Technology. 2Department of Computer Science and Technology, Harbin Institute of Technology, Shenzhen. 3School of Computer Science, China University of Geosciences. |
| Pseudocode | Yes | Algorithm 1 One Pass Late Fusion Multi-view Clustering |
| Open Source Code | No | The paper does not provide an explicit statement or link to the source code for the proposed OP-LFMVC methodology. It only mentions that implementations of other compared algorithms are publicly available. |
| Open Datasets | Yes | We conduct experimental comparison on a number of publicly available multi-view benchmark datasets, including 3Sources1, Football2, Olympics3, BBCSport4, Cal-205, Cora6, Citeseer7, SUNRGBD8. These dataset information is summarized in Table 1. As observed, the number of samples, kernels and categories of these datasets show considerable variation, providing a good platform to compare the performance of different clustering algorithms. 1http://mlg.ucd.ie/datasets/3sources.html 2http://mlg.ucd.ie/aggregation/ 3http://mlg.ucd.ie/aggregation/ 4http://mlg.ucd.ie/datasets/segment.html 5http://www.vision.caltech.edu/Image_ Datasets/Caltech101/ 6https://linqs-data.soe.ucsc.edu/public/ lbc/ 7https://linqs-data.soe.ucsc.edu/public/ lbc/ 8http://rgbd.cs.princeton.edu/ |
| Dataset Splits | No | The paper mentions repeating experiments but does not provide specific details on how the datasets were split into training, validation, and test sets (e.g., percentages, sample counts, or methodology for creating splits). |
| Hardware Specification | Yes | The experiments are conducted on a PC with Intel (R) Core (TM)-i9-10900X 3.7GHz CPU and 64G RAM in MATLAB environment. |
| Software Dependencies | No | The paper mentions the 'MATLAB environment' but does not specify a version number for MATLAB or any other software libraries or dependencies used with version numbers. |
| Experiment Setup | Yes | For all datasets, it is assumed that the true number of clusters k is known and set as the true number of classes. The clustering performance of all algorithms is evaluated by four widely used metrics: clustering accuracy (ACC), normalized mutual information (NMI), purity and rand index. For all compared algorithms, to alleviate the adverse influence of randomness by k-means, we repeat each experiment for 50 times and report the average values and the corresponding standard deviations. The proposed algorithm is free of hyper-parameter. However, among all compared algorithms, ONKC (Liu et al., 2017b), MKKM-Mi R (Liu et al., 2016b), LKAM (Li et al., 2016) and LF-MVC (Wang et al., 2019) have hyper-parameters to be tuned. By following the same way in literature, we reuse their released codes and tune the hyper-parameters by grid search to produce the best possible results on each dataset. |