Methods for Recovering Conditional Independence Graphs: A Survey
Authors: Harsh Shrivastava, Urszula Chajewska
JAIR 2024 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The focus of this paper is to review different methods and recent techniques developed to recover CI graphs. [...] Recovered graph structures for a sub-network of the E. coli consisting of 43 genes and 30 interactions with increasing number of samples. [...] CI graphs from u GLAD model used to analyse a lung cancer data from (Kaggle, 2022). [...] The CI graph recovered by u GLAD for the Infant Mortality 2015 data from CDC [...]. [left] u GLAD graph for archaea at family level in a collection of wastewater processing digesters. [...] [right] A dynamic network inference framework showing three snapshots of the automobile sensor network [...]. [left] Rectangular 10 20 latitude-longitude grids of windspeed locations [...]. [right] Graphical representation of latitude (left) and longitude factors [...]. The CI graph was recovered by the L2G algorithm, which is a deep unfolding approach to learn graph topologies, refer to (Pu et al., 2021). |
| Researcher Affiliation | Industry | Harsh Shrivastava EMAIL Urszula Chajewska EMAIL Microsoft Research, Redmond, USA |
| Pseudocode | No | The paper describes methods using mathematical formulations (e.g., Eq. 1, 2, 3, 4, 5) and architectural diagrams (Fig. 2) but does not include any explicitly labeled pseudocode or algorithm blocks. The procedural steps are described in narrative text without structured formatting. |
| Open Source Code | Yes | Table 1 lists some of the prominent methods for CI graph recovery along with their recommended implementations. This compilation will help the readers choose the right models for their applications. [...] GLAD https://github.com/Harshs27/GLAD (Shrivastava et al., 2020) [...] u GLAD https://github.com/Harshs27/u GLAD (Shrivastava et al., 2022b) [...] Tera Lasso https://github.com/kgreenewald/teralasso (Greenewald et al., 2019) [...] Sy Glasso https://github.com/ywa136/syglasso (Wang et al., 2020) [...] FASJEM https://github.com/QData/FASJEM (Wang et al., 2017) [...] ADMM https://github.com/tpetaja1/tvgl (Hallac, Park, Boyd, & Leskovec, 2017) [...] Newton-CG(MDMC) Matlab package (Zhang et al., 2018) |
| Open Datasets | Yes | CI graphs from u GLAD model used to analyse a lung cancer data from (Kaggle, 2022). Kaggle (2022). Lung Cancer. https://www.kaggle.com/datasets/nancyalaswad90/ lung-cancer?select=survey+lung+cancer.csv. [...] The CI graph recovered by u GLAD for the Infant Mortality 2015 data from CDC (United States Department of Health and Human Services, Division of Vital Statistics (DVS), 2015). United States Department of Health and Human Services, Division of Vital Statistics (DVS) (2015). Birth Cohort Linked Birth Infant Death Data Files, 2004-2015. Compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program, on CDC WONDER On-line Database. Accessed at https: //www.cdc.gov/nchs/data_access/vitalstatsonline.htm. |
| Dataset Splits | No | The paper describes the number of samples (M) and features (D) used in general terms for various methods (e.g., 'M samples and D features'). For instance, Figure 4 mentions 'M=10' and 'M=100' samples to illustrate the effect of sample size on FDR, TPR, and FPR. However, it does not explicitly provide any details about how datasets are split into training, validation, or test sets for reproducibility. |
| Hardware Specification | No | The paper describes various computational methods and algorithms, including those efficient for 'large scale problems involving thousands of variables'. However, it does not specify any particular hardware (e.g., GPU/CPU models, memory details) used for conducting experiments or for obtaining the results presented in its figures. |
| Software Dependencies | No | The paper mentions several software packages and libraries. For example, it notes a 'Graphical Lasso function implementation of python s scikit-learn package (Pedregosa et al., 2011)', a 'Python package' for G-ISTA, and 'R package' for JGL, GGMncv, and Miss Glasso. However, it does not provide specific version numbers for any of these software dependencies. |
| Experiment Setup | No | The paper discusses the theoretical underpinnings and general approaches of various CI graph recovery algorithms, such as the use of regularization parameters (e.g., 'λ Θ 1,off'). For the GLAD model, it mentions 'adaptive sequence of penalty hyperparameters'. However, it does not provide specific, concrete experimental setup details like learning rates, batch sizes, number of epochs, or specific optimizer configurations for any of the demonstrations or evaluations presented in the paper. |