Scalable Completion of Nonnegative Matrix with Separable Structure

Authors: Xiyu Yu, Wei Bian, Dacheng Tao

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To demonstrate the effectiveness of NMCSA for completing matrices with separable structures, we conduct empirical valuations both on synthetic and real datasets.
Researcher Affiliation Academia Xiyu Yu, Wei Bian, Dacheng Tao Center for Quantum Computation and Intelligent Systems, University of Technology Sydney
Pseudocode Yes Algorithm 1 Basis Selection by Random Projections; Algorithm 2 Scalable Optimisation for F and X
Open Source Code No The paper does not provide any information about open-source code availability or links to repositories for the described methodology.
Open Datasets Yes We further evaluate NMCSA on two real datasets, Jester1 and Movie Lens, for collaborative filtering. Both datasets are benchmarks and have been commonly used in the literature for matrix completions.
Dataset Splits No As no test data are available in these datasets, a common choice is to sample the available ratings by 50% for training and use the resting 50% for test (Wang et al. 2014; Aravkin et al. 2014). The paper explicitly mentions training and testing splits, but not a separate validation split.
Hardware Specification No All the experiments are performed in Matlab on a desktop computer. This statement is too general and does not specify any particular hardware components like CPU or GPU models.
Software Dependencies No All the experiments are performed in Matlab on a desktop computer. This only mentions 'Matlab' without a specific version number or other required software dependencies with versions.
Experiment Setup Yes Here, we fix the size of matrices to be 500 × 500, and vary the rank and sampling ratio in the following ranges, i.e., r ∈ {5, 8, 11, ..., 59} and ρ ∈ {0.01, 0.06, ..., 0.86}, according to (Wen, Yin, and Zhang 2012). Then, 50 independent matrix completion experiments are performed for each pair (r, ρ).