#Required packages

autograd 1.4 matplotlib 3.3.3 numpy 1.19.5 scipy 1.5.3 scikit-learn 1.0.2 panda s(with excel read) 1.1.4 tensorflow 2.4.0 causallearn 0.1.3.1 networkx 2.7.1 joblib 0.17.0

#List of helper files

-helper_bnlearn: Reads the bnlearn data, selects the adversarial parameters -helper_dag: Subrotiines relavent to sampling or estimatating dags -helper_data: Sachs dataset and trivariate dataset helper -helper_draw: Plots the loss functions

-helper_em_tf: Implements the WEM, and other EM related functions in tf -helper_mvn_tf: Subroutines relavent to mvn distribution in tf -helper_tf_model: Custom Keras model for the Algorithm 2

-helper_rs: Implements the local rejection sampling (RS) attack

-helper: Calls modeler algorithms (missDAG, missPC, etc) and calculates graph distance -helper_mvn: Subroutines relavent to mvn distribution -helper_em: Implements missDAG

-train_bnlear_adv: Trains LAMM for the bnlearn dataset -train_sachs_adv: Trains LAMM for the sachs dataset -train_trivariate: Trains LAMM for Gaussian SCM I

-load_bnlearn: Tests the LAMM for bnlearn -load_rs: Trains & tests the local rs attack on Gaussian SCM II. -load_sachs: Tests the LAMM for sachs dataset -load_trivariate: Tests the LAMM for Gassian SCM I

-External.notears: Contains the modified external notears package accessed at https://github.com/xunzheng/notears

#Sachs dataset requires downloading the supplementary files

1. Download https://www.science.org/doi/suppl/10.1126/science.1105809/suppl_file/sachs.som.datasets.zip
2. Put "1. cd3cd28.xls" to the project directory.

#Steps for testing the LAMM model:

1. Call the training file ex: train_sachs_adv.py,

2. Training curve is saved in a new folder starting with train_model_cfg_4_xx,

3. Call the corresponding test file ex: load_sachs.py, -You can specify the testing setting with parameters: bool_mcar and exp_type

4. It will output the results in the folder train_model_cfg_4_xx, -2_denea.png, summary_exp_type_1.csv: results for missPC, imputation -init_denea.png, summary.csv: results for missDAG

#Steps for testing the local RS attack:

1. Call load_rs.py -You can specify the testing setting with parameters: bool_mcar and exp_type

2. It will output the results in the folder train_model_cfg_4_xx, -2_denea.png, summary_exp_type_1.csv: results for missPC, imputation -init_denea.png, summary.csv: results for missDAG