Matrix Completion with Quantified Uncertainty through Low Rank Gaussian Copula

Authors: Yuxuan Zhao, Madeleine Udell

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results show the method yields state-of-the-art imputation accuracy across a wide range of data types, including those with high rank.
Researcher Affiliation Academia Yuxuan Zhao Cornell University yz2295@cornell.edu Madeleine Udell Cornell University udell@cornell.edu
Pseudocode Yes Algorithm 1 Imputation via low rank Gaussian copula fitting
Open Source Code No The paper does not provide an explicit statement about releasing code for the methodology or a link to a code repository.
Open Datasets Yes Movie Lens 1M dataset [20]
Dataset Splits Yes We use 80% of observation as training set, 10% as validation set, and 10% as test set, repeated 5 times.
Hardware Specification Yes On a laptop with Intel-i5-3.1GHz Core and 8 GB RAM
Software Dependencies No The paper mentions software like "R" and "julia" but does not provide specific version numbers for any software components.
Experiment Setup Yes We set n = 500 and p = 200. For continuous data, we use gj(z) = z to generate a low rank X = Z and gj(z) = z3 to generate a high rank X. We set k = 10, σ2 = 0.1 and the missing ratio as 40%. For 1-5 ordinal data and binary data, we use step functions gj with random selected cut points. We generate one X with high SNR σ2 = 0.1 and one X with low SNR σ2 = 0.5. We set k = 5 and the missing ratio as 60%. All experiments are repeated 20 times. ... LRGC (rank 10) takes 38 mins in R, soft Impute (rank 201) takes 93 mins in R, and GLRM-Bv S (rank 200) takes 25 mins in julia.