gaotianxiang / A-star-lasso

A * Lasso

Python implementation of paper A* Lasso for Learning a Sparse Bayesian Network Structure for Continuous Variables

Generating simulated data

python make_data.py

This script will generate sample data. You can adjust the number of datasets you want to simulate as well as the strength of your edges (beta values). Sample benchmark networks are included in the networks/ folder. You can also specify your own networks by editing these files.

Running DP Lasso

All the code for running DP Lasso is in the dp_lasso/ folder. There is a demo called run_dp_lasso.py that you should run. DP Lasso generates the optimal solution. However, it requires exponential time.

python run_dp_lasso.py

Running A* Lasso

All the code for running A* Lasso is in the a_star_lasso/ folder. There is a demo called run_astar_lasso.py that you should run. In the demo, you have the option to run:

• A* Lasso - using wrap_a_star_lasso.py This runs the optimal A* Lasso.

python run_astar_lasso.py

• A* Lasso with Pruning - using wrapAstarLassoPrune.m
  This uses our heuristic strategy to prune the queue so that the solution is attained much faster without significant compromise in quality. You should use this if the networks are larger than 20 nodes.