Large-Scale Multi-View Spectral Clustering via Bipartite Graph

Authors: Yeqing Li, Feiping Nie, Heng Huang, Junzhou Huang

AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experimental results on five benchmark datasets demonstrate the effectiveness and efficiency of the proposed method, where our method runs up to nearly 3000 times faster than the state-of-the-art methods.
Researcher Affiliation Academia University of Texas at Arlington, Arlington, TX, 76019, USA {yeqing.li@mavs.uta.edu, feipingnie@gmail.com, heng@uta.edu, jzhuang@uta.edu}
Pseudocode Yes Algorithm 1 Multi-view Spectral Clustering (MVSC) 1: Input: Data matrix of all views X(v) Rn d(v) for v 1 . . . V , Number of classes K, Number of salient points m, parameter r. 2: Output: Cluster labels Y of each data points, all salient points U and cluster labels of all salient points. 3: Generate m salient points using k-means on concatenate features; 4: Compute affinity matrix Z(v) of each view. 5: Compute Laplacian L(v) of each view; 6: Initialize a(v) = 1/K; 7: repeat 8: Compute G by using Eq. (14); 9: Update a(v) by using Eq. (8); 10: until Converges. 11: Treat each row of G as new representation of each data point and compute the clustering labels Y by using kmeans algorithm.
Open Source Code No The paper does not provide an explicit statement about releasing the source code for the described methodology or a link to a code repository.
Open Datasets Yes We conduct several experiments to evaluate the performance of the proposed methods on five benchmarks datasets. These datasets are summarized in Table 2. (Table 2 lists datasets like HW, Caltech-101 (Fei-Fei, Fergus, and Perona 2007), Reuters, NUS-WIDE-Object (Chua et al. July 8 10 2009), Animal with attributes (AWA), with associated URLs like https://archive.ics.uci.edu/ml/datasets/Multiple+Features for HW and http://lms.comp.nus.edu.sg/research/NUS-WIDE.htm for NUS.)
Dataset Splits Yes Five fold cross-validation is used and we report the mean purity, the mean NMI and the mean testing time.
Hardware Specification Yes All our experiments are conducted on a desktop computer with a 3.4GHz Intel Core i7 CPU and 12GB RAM, Mat Lab 2012a (64bit).
Software Dependencies Yes Mat Lab 2012a (64bit).
Experiment Setup Yes We search the parameter r in logarithm form (log10 r from 0.1 to 2 with step size 0.2. We also set m = 400 and construct 8-nearest-neighbour graph between raw All the experiments are repeated for 10 times and average results are reported.