Ordinal Graphical Models: A Tale of Two Approaches
Authors: Arun Sai Suggala, Eunho Yang, Pradeep Ravikumar
ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section we present the performance of Consecutive Ratio model (Consec model), and our estimator for probit model (Probit Direct) on various synthetic and real world datasets. |
| Researcher Affiliation | Collaboration | Arun Sai Suggala 1 Eunho Yang 2 3 Pradeep Ravikumar 1 ... 1Carnegie Mellon University, Pittsburgh, USA 2School of Computing, KAIST, Daejeon, South Korea 3AITrics, Seoul, South Korea. |
| Pseudocode | No | The paper describes methods through prose and mathematical equations but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | The Health Information National Trends Survey (HINTS) is a nationally representative survey conducted by the National Cancer Institute (NCI) every few years. In this work we analyze HINTS-FDA data which is a special data collected by NCI in partnership with the Food and Drug Administration (FDA) and is made publicly available by NCI. |
| Dataset Splits | Yes | The best tuning parameter for all the estimators described above is selected using 5 fold cross validation. For Probit EM, Discrete model, Consec model and Oracle we use the standard cross validation technique where we pick the best tuning parameter based on log likelihood on validation set. For Probit Direct we use the following k-fold cross validation technique. We partition the data set into k subsets. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running the experiments. |
| Software Dependencies | No | The paper mentions software tools and estimators like 'glasso' (Friedman et al., 2008), 'Probit EM+MCMC based estimator' (Guo et al., 2015), and 'pseudo-likelihood based estimator' (Jalali et al., 2011), but it does not specify any version numbers for these or other software dependencies. |
| Experiment Setup | Yes | In all our experiments we fix the number of nodes in the graph to 50 and set number of categories for each ordinal variable to 5. To reduce the variance, we average results over 10 repetitions. ... The inverse covariance matrix of the latent variables is chosen as follows: !|j k| if |j k| 1 0 otherwise . We pick an ! 2 ( 1, 1) in our experiments and set the thresholds ( ) at node j as : (j) = [ Inf, 10, 0.7, 0.7, 10, Inf]. |