Joint Structural Estimation of Multiple Graphical Models
Authors: Jing Ma, George Michailidis
JMLR 2016 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We present three simulation studies to evaluate the performance of JSEM. Other methods compared include the separate estimation method Glasso, where the Graphical lasso by Friedman et al. (2008) is applied to each graphical model separately, joint estimation by Guo et al. (2011), denoted by JEM-G, the Group Graphical Lasso denoted by GGL by Danaher et al. (2014), and the structural pursuit method MGGM by Zhu et al. (2014). [...] Applications to a climate data set and a breast cancer data set are also discussed. |
| Researcher Affiliation | Academia | Jing Ma EMAIL Department of Biostatistics and Epidemiology Perelman School of Medicine University of Pennsylvania [...] George Michailidis EMAIL Department of Statistics University of Florida |
| Pseudocode | No | The paper describes the Joint Structural Estimation Method (JSEM) with a two-step procedure, detailing the mathematical formulations for each step. However, it does not present these steps within a clearly labeled 'Pseudocode' or 'Algorithm' block, nor does it use code-like structured formatting for the procedure. |
| Open Source Code | No | In this work, we use the R-package grpreg (Breheny and Huang, 2009) for implementation of the group lasso penalized optimization (2) and the glasso (Friedman et al., 2008) one for solving (4). |
| Open Datasets | Yes | The data used in this study are monthly measurements from January 2001 to June 2005 on 16 variables [...] from CRU (http://www.cru.uea.ac.uk/cru/data), NOAA (http://www.esrl.noaa.gov/gmd/dv/ftpdata.html), NASA (http://disc.sci.gsfc.nasa.gov/aerosols) and NCDC (ftp://ftp.ncdc.noaa.gov/pub/data/nsrdb-solar/). [...] The breast cancer data set (TCGA, 2012) contains RNA-seq measurements for 17296 genes from 1033 breast cancer specimens... |
| Dataset Splits | Yes | To perform stability selection, we ran our method 50 times on two randomly drawn complementary pairs of sizes 8 and 9, and kept only edges that are selected over 70% of the time. [...] For each of the three methods considered here, we used BIC on the normalized data to select the optimal tuning parameters and coupled each method with complementary pairs stability selection (Shah and Samworth, 2013) to infer the related climate networks. |
| Hardware Specification | No | The paper discusses the computational complexity of the method, such as O(Kp3) and O(K2p2), and mentions the use of R packages for implementation. However, it does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used to run the experiments or simulations. |
| Software Dependencies | No | In this work, we use the R-package grpreg (Breheny and Huang, 2009) for implementation of the group lasso penalized optimization (2) and the glasso (Friedman et al., 2008) one for solving (4). |
| Experiment Setup | Yes | We recommend choosing the tuning parameters via the Bayesian information criterion (BIC). Specifically, for a given λ, we define BIC for the proposed method as [...] The optimal tuning parameter is thus λ = argminλ Dn BIC(λ), where the set of values Dn is chosen such that for every λj Dn (nk = n): λj = cj |gmax| + p log G0 n, cj = 0.02 j, j = 1, . . . , 20. |