A Completely Tuning-Free and Robust Approach to Sparse Precision Matrix Estimation
Authors: Chau Tran, Guo Yu
ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Through comprehensive numerical studies, our methods demonstrate favorable statistical performance. Remarkably, our methods exhibit strong robustness to the violation of the Gaussian assumption and significantly outperform competing methods in the heavy-tailed settings. |
| Researcher Affiliation | Academia | 1Department of Statistics and Applied Probability, University of California, Santa Barbara, CA, USA. |
| Pseudocode | Yes | Algorithm 1 Simulate λj, for j = 1, . . . , d |
| Open Source Code | No | The paper states that "All numerical experiments are implemented in R (R Core Team, 2021). The CLIME estimator is computed using the R package flare (Li et al., 2015); the TIGER and GLasso estimators are computed using the R package huge (Zhao et al., 2012)." This indicates the use of third-party software, not the release of the authors' own source code for the proposed methods. |
| Open Datasets | Yes | We apply our proposed methods to reconstruct the interaction network from human gene expression data in the BDgraph R package (Mohammadi & Wit, 2019), which was previously studied by Bhadra & Mallick (2013); Mohammadi & Wit (2015); Liu & Wang (2017). This dataset consists of n = 60 individuals of Northern and Western European ancestry from Utah, whose genotypes are available online at the Sanger Institute website1. |
| Dataset Splits | No | The paper mentions "For CLIME and GLasso, the optimal tuning parameter values are chosen using a validation set approach." but does not provide specific details on the dataset splits (e.g., percentages, sample counts) for training, validation, or testing, nor does it specify if these splits are publicly available or how they were created in a reproducible manner. |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments. |
| Software Dependencies | Yes | All numerical experiments are implemented in R (R Core Team, 2021). The CLIME estimator is computed using the R package flare (Li et al., 2015); the TIGER and GLasso estimators are computed using the R package huge (Zhao et al., 2012). |
| Experiment Setup | Yes | For each type of graph, we set n = 100 and d {25, 50, 100, 200, 400} and repeat the simulation 50 times. For CLIME and GLasso, the optimal tuning parameter values are chosen using a validation set approach... We use the regularization parameter λ = p(log d)/n for TIGER as suggested in Liu & Wang (2017)... While g Rank Lasso is completely tuning-free, g Rank MCP requires some light tuning. We use the HBIC (Wang et al., 2020) to select the best value of η in (5). |