Multi-view Sparse Co-clustering via Proximal Alternating Linearized Minimization

Authors: Jiangwen Sun, Jin Lu, Tingyang Xu, Jinbo Bi

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

Reproducibility Variable Result LLM Response
Research Type Experimental We implemented our approach using Matlab and validated it first on synthetic data that was simulated with known row and column clusters. ... Then we evaluated our approach on two benchmark datasets with known subject clusters but unknown feature clusters. ... Table 1 provides the NMI values. ... Table 3 summarizes the NMI values of all methods on the benchmark datasets.
Researcher Affiliation Academia Jiangwen Sun JAVON@ENGR.UCONN.EDU Jin Lu JIN.LU@ENGR.UCONN.EDU Tingyang Xu TIX11001@ENGR.UCONN.EDU Jinbo Bi JINBO@ENGR.UCONN.EDU Department of Computer Science and Engineering, University of Connecticut, Storrs, CT 06269 USA
Pseudocode Yes Algorithm 1 Multi-view rank one matrix approximation
Open Source Code No The paper does not provide an explicit statement about the release of source code or a link to a code repository for the methodology described.
Open Datasets Yes UCI Handwritten digits dataset: We downloaded the handwritten digits data from the UCI repository. Crowd-sourcing dataset: This dataset was downloaded from a study of Crowd-sourcing Big Data (http://web.eecs.umich.edu/~mozafari/projects.html) (Mozafari et al., 2012).
Dataset Splits No The paper mentions running 'multiple trials (i.e., each trial used randomly 80% of the data)' but does not explicitly define standard train/validation/test dataset splits with specific percentages or sample counts, nor does it refer to predefined splits or cross-validation for model selection/evaluation.
Hardware Specification No The paper does not provide specific hardware details such as CPU/GPU models, memory amounts, or cloud computing specifications used for running experiments.
Software Dependencies No The paper states, 'We implemented our approach using Matlab', but it does not specify a version number for Matlab or any other software dependencies.
Experiment Setup Yes Empirically, we observed good performance when the parameter sv is set to the number of selected features whose percentage of accumulated latent in principle component analysis (PCA) is over 90%. ... We found that if we set the initial vector v proportionally with the first moment of PCA, our method was able to perform better than those with the initial vectors of all-ones or random-ones.