Multi-View Clustering in Latent Embedding Space

Authors: Man-Sheng Chen, Ling Huang, Chang-Dong Wang, Dong Huang3513-3520

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments conducted on several real-world multi-view datasets have demonstrated the superiority of our approach. In this section, extensive experiments are conducted to validate the superiority of the proposed method.
Researcher Affiliation Academia Man-Sheng Chen,1,2 Ling Huang,1,2 Chang-Dong Wang,1,2 Dong Huang3 1School of Data and Computer Science, Sun Yat-sen University, Guangzhou, China 2Guangdong Province Key Laboratory of Computational Science, Guangzhou, China 3College of Mathematics and Informatics, South China Agricultural University, Guangzhou, China
Pseudocode Yes Algorithm 1 MCLES Input: Multi-view matrices X = X(1), ..., X(V ) , cluster number c, parameters α, β and γ, and the embedding dimension d of latent representation. Initialize: W = 0, S = 0 and r = 1; Initialize H and P with random values. 1: repeat 2: repeat 3: r r + 1; 4: Update W according to Eq. (12). 5: until convergence 6: Update H according to Eq. (14). 7: For each i, update the i-th column of S by solving the problem in Eq. (18). 8: Update P, which is formed by the c eigenvectors of Ls = D ST +S 2 corresponding to the c smallest eigenvalues. 9: until convergence Output: W, H, S and P.
Open Source Code Yes The code of our method is available on the github1. 1https://github.com/Ttuo123/MCLES
Open Datasets Yes Datasets Description Yale2: It is a widely used face image dataset consisting of 165 gray-scale images belonging to 15 distinct subjects... 2http://cvc.yale.edu/projects/yalefaces/yalefaces.html. MSRCv1 (Winn and Jojic 2005): It is an image dataset... ORL3: It is a widely used face image dataset... 3http://www.cl.cam.ac.uk/research/dtg/. BBCSport (Xia et al. 2014): It is a document dataset...
Dataset Splits No The paper describes parameter analysis and running experiments multiple times ('run 20 times for each experiment'), but does not explicitly provide specific training/validation/test dataset splits (e.g., percentages or sample counts) or a detailed cross-validation setup.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments, such as GPU models, CPU types, or memory specifications.
Software Dependencies No The paper mentions algorithms used (e.g., ADMM) and general tools ('Many existing quadratic programming packages'), but does not list specific software dependencies with version numbers.
Experiment Setup Yes Table 1: The default values of the four parameters. Parameter Yale MSRCv1 ORL BBCSport d 30 70 50 40 α 0.8 0.8 0.8 0.8 β 0.4 0.4 0.5 2 γ 0.004 0.004 0.004 0.004