A Constraint Optimzation Approach to Causal Discovery from Subsampled Time Series Data

This page will provide code and benchmarks as supplementary material to the work.

A Constraint Optimization Approach to Causal Discovery from Subsampled Time Series Data
Antti Hyttinen, Sergey M. Plis, Matti Järvisalo, Frederick Eberhardt, and David Danks, International Journal of Approximate Reasoning 2017.

Causal Discovery from Subsampled Time Series Data by Constraint Optimization
Antti Hyttinen, Sergey M. Plis, Matti Järvisalo, Frederick Eberhardt, and David Danks, International Conference on Probabilistic Graphical Models 2016.

Code

The code package including R code for the simulations and analysis, ASP codes for clingo/clasp and perl code for producing cnf/wcnf encodings from the ASP encodings. Answer Set Programming encodings proposed in the paper is available here. Last updated 27.6.2018.

Benchmarks

Here we also provide benchmarks appearing in Figures 12 and 13 in the paper. These synthetic instances.

Figure 12 CNFs and ASP files are here. (for ASP and SAT-solvers, task 1) Remember to input supersample.pl to clingo in addition to the data file.

Figure 13 wCNFs and ASP files are here. (for ASP and MaxSAT-solvers, task 2) Remember to input supersample_weighted.pl to clingo in addition to the data dile.

Solvers

Solvers used in the paper:

  • Clingo ASP solver (for Tasks 1 and 2)
  • Lingeling SAT solver (for Task 1)
  • Glucose SAT solver (for Task 1)
  • QMaxSAT MaxSAT solver (for Task 2)
  • LMHS MaxSAT solver (for Task 2)
  • PrimalDual MaxSAT solver (for Task 2)
  • Eva500a MaxSAT solver (for Task 2)
  • Open-WBO MaxSAT solver (for Task 2)
  • MSCG MaxSAT solver (for Task 2)