Optimization Methods for Sparse Pseudo-Likelihood Graphical Model Selection

Authors: Sang Oh, Onkar Dalal, Kshitij Khare, Bala Rajaratnam

NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this paper, we address this crucial gap by proposing two proximal gradient methods (CONCORDISTA and CONCORD-FISTA) for performing ℓ1-regularized inverse covariance matrix estimation in the pseudo-likelihood framework. We present timing comparisons with coordinate-wise minimization and demonstrate that our approach yields tremendous payoffs for ℓ1-penalized partial correlation graph estimation outside the Gaussian setting, thus yielding the fastest and most scalable approach for such problems.
Researcher Affiliation Academia Sang-Yun Oh Computational Research Division Lawrence Berkeley National Lab syoh@lbl.gov Onkar Dalal Stanford University onkar@alumni.stanford.edu Kshitij Khare Department of Statistics University of Florida kdkhare@stat.ufl.edu Bala Rajaratnam Department of Statistics Stanford University brajarat@stanford.edu
Pseudocode Yes Algorithm 1 CONCORD-ISTA Algorithm 2 CONCORD-FISTA
Open Source Code No The paper mentions 'gconcord package for R available at http://cran.r-project.org/web/packages/gconcord/' which refers to a baseline implementation. It also states 'Algorithms 1 and 2 were also written in C++ then interfaced to R for testing,' but there is no explicit statement or link indicating that the authors' own source code for CONCORD-ISTA/FISTA is made publicly available.
Open Datasets Yes In this section, the performance of proposed methods are assessed on a breast cancer dataset [12]. This dataset contains expression levels of 24481 genes on 266 patients with breast cancer.
Dataset Splits No The paper does not provide specific training, validation, or test dataset splits (e.g., percentages or sample counts). It mentions using synthetic and real datasets, and initial guess/convergence criteria, but not how data was partitioned for training/validation/testing.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, memory, or cloud instance types used for running the experiments.
Software Dependencies Yes For this work, we opted to using Eigen library [11] for its sparse linear algebra routines written in C++. Algorithms 1 and 2 were also written in C++ then interfaced to R for testing.
Experiment Setup Yes The initial guess, Ω(0), and the convergence criteria was matched to those of coordinate-wise CONCORD implementation. The line search for iteration k starts with an initial step size τ(k,0) and reduces the step with a constant factor c until the new iterate satisfies the sufficient descent condition. Figure 1: Convergence of CONCORD-ISTA and CONCORD-FISTA for threshold subg < 10 5. A choice of p-value < 0.03 yields a reduced dataset with p = 4433 genes.