A Graphical Lasso Benchmark

Benchopt is a package to simplify and make more transparent and reproducible comparisons of optimization methods. This benchmark is dedicated to solvers of the Graphical Lasso estimator (Banerjee et al., 2008):

$$\min_{\Theta \succ 0} - \log \det (\Theta) + \langle \Theta, S \rangle + \alpha \Vert \Theta \Vert_{1,\mathrm{off}},$$

where $\Theta$ is the optimization variable, $S$ is the empirical covariance matrix and $\alpha$ is the regularization hyperparameter.

Install

This benchmark can be run using the following commands, which first create a dedicated Conda environment:

$ conda create -n glasso_bench_env python=3.10
$ conda activate glasso_bench_env
$ pip install -U benchopt
$ git clone https://github.com/Perceptronium/benchmark_graphical_lasso
$ pip install gglasso
$ git clone https://github.com/skggm/skggm ./benchmark_graphical_lasso/benchmark_utils/skggm
$ pip install Cython
$ pip install -e ./benchmark_graphical_lasso/benchmark_utils/skggm/
$ benchopt run ./benchmark_graphical_lasso --config ./benchmark_graphical_lasso/simple_conf.yml

Please visit https://benchopt.github.io/ for more details on using the benchopt ecosystem.