Robust Self-Weighted Multi-View Projection Clustering
Authors: Beilei Wang, Yun Xiao, Zhihui Li, Xuanhong Wang, Xiaojiang Chen, Dingyi Fang6110-6117
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experimental results on different synthetic datasets and real-world datasets demonstrate that the proposed algorithm outperforms other state-ofthe-art methods on clustering performance and robustness. |
| Researcher Affiliation | Academia | 1School of Information Science and Technology, Northwest University, Xi an 710127, P.R. China 2School of Information Science and Engineering, Shandong Normal University, Jinan 250358, P.R. China 3School of Computer Science and Engineering, University of New South Wales, Sydney, NSW 2052, Australia 4School of Communications and Information Engineering, Xi an University of Posts & Telecommunications, Xi an 710121, P.R. China |
| Pseudocode | Yes | Algorithm 1 Optimization in problem (7) Input: X = {X1, X2, ..., Xv} , Xv Rdv n, clustering number k, parameter β, η and γ. Output: affinity matrix S Rn n with exact c connected components, where c = k; projection matrix W = {W1, W2, ..., Wv} , W Rdv d v. Initialize each vector si of S by the optimal solution to the following problem, where the initial value of αv is 1 min s T i 1=1,0 sij 1 v xv i xv j 2 2sij + βs2 ij) Initialize Wv by the optimal solution to the following problem: min Wv Tr(W T v Xv LSXT v Wv), s.t. W T v Xv XT v Wv = I repeat i. update αv by solving the problem (6), ii. update F by solving the problem (11), iii. update the projection matrix Wv for each view by solving the problem (18), iv. update each row vector of S by solving the problem (23). until converge |
| Open Source Code | No | The paper does not provide any statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | In Table 1, we briefly summarize the datasets used in this paper, namely MSRC-v1 (Winn and Jojic 2005), Caltech101 (Fei-Fei, Fergus, and Perona 2007) (here, we use two regular subsets of Caltech101-7 and Caltech101-20), Handwritten numerals (HW) (Asuncion and Newman 2007), NUSWIDE (Chua et al. 2009) (We select 12 categories of animal images, and the first 200 images are selected for each category), ORL Face (Samaria and Harter 1994) datasets. |
| Dataset Splits | No | The paper does not explicitly provide training/validation/test dataset splits, percentages, or absolute sample counts. It discusses using various datasets for evaluation but does not detail how they were partitioned for training or validation. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used to run the experiments. |
| Software Dependencies | No | The paper does not list specific software dependencies with version numbers. |
| Experiment Setup | Yes | For our proposed algorithm, except that the neighbor numbers K of the MSRC-v1 and ORL Face datasets are set to 30 and 5, respectively, the rest is 15. The parameter γ is searched between 1e 6 to 1e6, and the step size is 0.5. In addition, we search for the dimensionality d corresponding to the best clustering result by searching all the dimensionalities of each view in the original dataset. For the sake of simplicity, we set the step size of the parameter γ to 3. For other algorithms, we set the parameters to be optimal. All the experiments are repeated for 20 times to obtain the mean and standard deviation of clustering results. |