Centerless Multi-View K-means Based on the Adjacency Matrix

Authors: Han Lu, Quanxue Gao, Qianqian Wang, Ming Yang, Wei Xia

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments We evaluate our model on a toy dataset and six benchmark datasets through some experiments implemented on a Windows 10 desktop computer with a 2.40GHz Intel Xeon Gold 6240R CPU, 64 GB RAM, and MATLAB R2021a (64-bit).Experimental Results We do experiments on six datasets and present the metrics comparison of the above methods in Tables. 1, 2, and 3, and specially for SC (best), only the best results among each view are recorded that individually applying single view spectral clustering on.
Researcher Affiliation Academia 1 School of Telecommunications Engineering, Xidian University, Xi an 710071, P.R.China. 2 Mathematics department of the University of Evansville, Evansville, IN 47722 USA
Pseudocode Yes Algorithm 1: Solving the Model (7)
Open Source Code Yes Code and data are available1. 1https://github.com/luhan0/CMKOA
Open Datasets Yes MSRC (Winn and Jojic 2005) includes 7 kinds of objects with 210 images. ORL2 includes 400 pictures of 40 people. 2http://www.cl.cam.ac.uk/research/dtg/attarchive/facedatabase. html HW (Dua and Graff 2019) includes 10 digits with 2,000 images generated from UCI machine learning repository. Mnist4 (Deng 2012) includes 4 categories of handwritten digits, i.e., from digit 0 to 3, with 4,000 images. Reuters (Apt e, Damerau, and Weiss 1994) includes 6 categories with 18758 documents. NUS-WIDE (Chua et al. 2009) includes 31 categories with 30000 object images
Dataset Splits No The paper does not explicitly provide details about training, validation, or test dataset splits, nor does it refer to specific file names or URLs for custom splits.
Hardware Specification Yes implemented on a Windows 10 desktop computer with a 2.40GHz Intel Xeon Gold 6240R CPU, 64 GB RAM, and MATLAB R2021a (64-bit).
Software Dependencies Yes MATLAB R2021a (64-bit)
Experiment Setup Yes We set the anchor rate θ = 0.5 on MSRC, ORL, Mnist4 and HW, θ = 0.1 on Reuters and θ = 0.02 on NUS-WIDE. Initialize: λ, Ω, r, p, J, H(v), Y(v), αv = 1/V, (v = 1 . . . V ), ρ = 1.1, µ = 10^-4, µmax = 10^10.