Causal Learner: A Toolbox for Causal Structure and Markov Blanket Learning
If you find this toolbox useful, please consider citing:
Zhaolong Ling, Kui Yu, Yiwen Zhang, Lin Liu, and Jiuyong Li. Causal Learner: A Toolbox for Causal Structure and Markov Blanket Learning[J]. Pattern Recognition Letters, 2022, 163: 92-95.
@article{ling2022causal,
title={Causal Learner: A Toolbox for Causal Structure and Markov Blanket Learning},
author={Ling, Zhaolong and Yu, Kui and Zhang, Yiwen and Liu, Lin and Li, Jiuyong},
journal = {Pattern Recognition Letters},
volume = {163},
pages = {92-95},
year = {2022},
issn = {0167-8655},
}
Causal Learner is a toolbox for learning causal structure and Markov blanket (MB) from data. It integrates functions for generating simulated Bayesian network data, a set of state-of-the-art global causal structure learning algorithms, a set of state-of-the-art local causal structure learning algorithms, a set of state-of-the-art MB learning algorithms, and functions for evaluating algorithms. The data generation part of Causal Learner is written in R, and the rest of Causal Learner is written in MATLAB. Causal Learner aims to provide researchers and practitioners with an open-source platform for causal learning from data and for the development and evaluation of new causal learning algorithms.
********** Names of the 31 Bayesian networks are as follows **********
(1) Discrete Bayesian Networks
Small Networks (<20 nodes) asia, cancer, earthquake, sachs, survey
Medium Networks (20–50 nodes) alarm, barley, child, insurance, mildew, water
Large Networks (50–100 nodes) hailfinder, hepar2, win95pts
Very Large Networks (100–1000 nodes) andes, diabetes, link, pathfinder, pigs, munin1, munin2, munin3, munin4';
Massive Networks (>1000 nodes) munin
(2) Continues Bayesian Networks (Gaussian)
Medium Networks (20–50 nodes) ecoli70, magic-niab
Large Networks (50–100 nodes) magic-irri
Very Large Networks (101–1000 nodes) arth150
(3) Continues Bayesian Networks (Conditional Linear Gaussian)
Small Networks (<20 nodes) sangiovese
Medium Networks (20–50 nodes) mehra-original, mehra-complete
********** Names of the 26 algorithms are as follows **********
(1) global causal structure learning: GSBN, GES, PC, MMHC, PCstable, F2SL_c, F2SL_s
(2) local causal structure learning: PCD_by_PCD, MB_by_MB, CMB, LCS_FS
(3) Markov blanket learning: GS, IAMB, interIAMB, IAMBnPC, interIAMBnPC, FastIAMB, FBED, MMMB, HITONMB, PCMB, IPCMB, MBOR, STMB, BAMB, EEMB, MBFS
********** Names of the 13 Metrics evaluating algorithms are as follows **********
(1) global causal structure learning:
----- In accuracy: Ar_F1, Ar_precision, Ar_recall, SHD, Miss, Extra, Reverse
----- In efficiency: running time
(2) local causal structure learning:
----- In accuracy: Ar_F1, Ar_precision, Ar_recall, SHD, Miss, Extra, Reverse, Undirected
----- In efficiency: running time, number of conditional independence tests
(3) Markov blanket learning:
----- In accuracy: Adj_F1, Adj_precision, Adj_recall
----- In efficiency: running time, number of conditional independence tests