07.12.2018  Regular Paper  Ausgabe 3/2019 Open Access
Telling cause from effect by local and global regression
 Zeitschrift:
 Knowledge and Information Systems > Ausgabe 3/2019
Publisher's Note
1 Introduction
2 Preliminaries
2.1 Kolmogorov complexity
2.2 Minimum Description Length principle
3 Information theoretic causal inference
3.1 Causal inference by Kolmogorov complexity
3.2 Causal inference by MDL
3.2.1 Intuition of the conditional encoding
3.2.2 Complexity of the marginals
3.2.3 Complexity of the conditional model
3.2.4 Complexity of the conditional data
3.2.5 Complexity of the conditional
4 Identifiability and significance
4.1 Identifiability
4.2 Significance by hypercompression
4.3 Significance by confidence
5 The Slope algorithm
5.1 Calculating the conditional scores
5.2 Causal direction and confidence
5.3 Combining basis functions
5.4 Computational complexity
6 Related work
7 Experiments
7.1 Evaluation measures
7.2 Synthetic data
7.2.1 Accuracy
7.2.2 Confidence
7.3 Identifiability of ANMs on synthetic data
7.3.1 Nondeterminacy
7.3.2 GP simulated data
7.4 Realworld data
7.4.1 Accuracy curves and overall accuracy
7.5 Runtime
7.6 Case study: Octet binary semiconductors
8 Discussion
9 Conclusion
Acknowledgements
Appendix
Slope

Sloper

Cure

Resit

IGCI

ANM
 

Tübingen\(_{98}\)  
\(\text {ROC}_X\) 
0.898
 0.865  0.424  0.573  0.671  0.472 
\(\text {ROC}_Y\) 
0.897
 0.862  0.413  0.564  0.675  0.472 
\(\text {PR}_X\) 
0.962
 0.948  0.716  0.791  0.808  0.734 
\(\text {PR}_Y\) 
0.728
 0.705  0.232  0.265  0.600  0.255 
AUAC 
0.942
 0.927  0.588  0.676  0.736  0.713 
Tübingen\(_{79}\)  
\(\text {ROC}_X\) 
0.812
 0.792  0.381  0.508  0.388  0.469 
\(\text {ROC}_Y\) 
0.851
 0.830  0.414  0.528  0.422  0.502 
\(\text {PR}_X\) 
0.942
 0.935  0.740  0.800  0.675  0.742 
\(\text {PR}_Y\) 
0.575
 0.573  0.200  0.254  0.269  0.232 
AUAC 
0.933
 0.924  0.819  0.534  0.715  0.802 