A Pairwise Pseudo-likelihood Approach for Matrix Completion with Informative Missingness

Authors: Jiangyuan Li, Jiayi Wang, Raymond K. W. Wong, Kwun Chuen Gary Chan

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The efficacy of our method is validated via numerical experiments, positioning it as a robust tool for matrix completion to mitigate data bias.
Researcher Affiliation Academia Jiangyuan Li* Department of Statistics Texas A&M University College Station, TX 77843 jiangyuanli@tamu.edu Jiayi Wang* Department of Mathematical Sciences University of Texas at Dallas Richardson, TX 75080 jiayi.wang2@utdallas.edu Raymond K. W. Wong Department of Statistics Texas A&M University College Station, TX 77843 raywong@tamu.edu Kwun Chuen Gary Chan Department of Biostatistics University of Washington Seattle, WA 98195 kcgchan@uw.edu
Pseudocode Yes Algorithm 1 Projected gradient descent Initialize: Initialize A(0) randomly, set learning rate η. for t = 0 to T do K(t) = A(t) η ℓ(A(t)) Q(t) = Sλ(K(t)) A(t + 1) = POCS(Q(t)) end for Algorithm 2 POCS Initialize: Input matrix Q Rm1 m2. t = 0. while Q = Q do Q = Q Q = Q 1 m1m2 Pm1 i=1 Pm2 j=1 Qi,j J Q i,j = 1 | Qi,j| α Qi,j + 1 | Qi,j| > α sign Qi,j α end while
Open Source Code Yes The code is publicly available on Git Hub2. 2https://github.com/jiangyuan-li/mc-w-pseudolikelihood
Open Datasets Yes Tobacco Dataset. This dataset is available in Table 11 in [9]... Coat Shopping Dataset. This dataset is available at https://www.cs.cornell.edu/~schnabts/mnar... Yahoo! Webscope Dataset. This dataset is available at https://webscope.sandbox.yahoo.com/catalog.php?datatype= r&did=3...
Dataset Splits Yes We use the observed entries as training data and equally split the unobserved data as validation and test data. The validation data is used for hyper-parameter tuning in each method.
Hardware Specification No The paper mentions "advanced computing resources provided by Texas A&M High Performance Research Computing" in the acknowledgements, but does not specify any particular GPU models, CPU types, or other detailed hardware specifications used for the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers for replication.
Experiment Setup Yes Since the objective function is convex in the proposed method, we only tune the regularization parameter λ, and fix the number of iterations as T = 100 and step size η = 1.0 in Algorithm 1.