GOBNILP (GOBNILP)

Module name	gobnilp
Package	GOBNILP
Version	#4347c64
Language	C
Docs	here
Paper	Cussens[1], Bartlett and Cussens[2], Cussens et al.[3], Cussens[4]
Graph type	DAG
MCMC	No
Edge constraints	Yes
Data type	C, D
Data missingness
Intervention type
Docker	bpimages/gobnilp:4347c64

Globally optimal Bayesian network learning using integer linear programming

GOBNILP is a score based method using integer linear programming (ILP) for learning an optimal DAG for a Bayesian network with limit on the maximal number of parents for each node. It is a two-stage approach where candidate parent sets for each node are discovered in the first phase and the optimal sets are determined in a second phase.

Important

This module only works on the AMD64 architecture.

Some fields described

edgeConstraints Name of the JSON file containing background knowledge
extra_args File with extra arguments to pass to the solver. The file should be placed in resources/extra_args and the format is the same as used by gobnilp (see the docs).
gap_limit Gap limit.
time_limit Time limit in seconds for the solver (not including the time to build the score tables).
timeout Use the best DAG found so far after this number of seconds.

Example JSON

[
  {
    "id": "gobnilp-bge",
    "continuous": true,
    "score_type": "BGe",
    "extra_args": null,
    "plot": false,
    "palim": 3,
    "alpha_mu": [
      1e-05,
      0.0001,
      0.001
    ],
    "alpha_omega_minus_nvars": 2,
    "alpha": null,
    "time_limit": null,
    "gap_limit": null,
    "prune": true,
    "timeout": 800,
    "edgeConstraints": "edgeConstraints.json"
  },
  {
    "id": "gobnilp-bde",
    "continuous": false,
    "score_type": "BDeu",
    "extra_args": null,
    "plot": false,
    "palim": 4,
    "alpha_mu": null,
    "alpha_omega_minus_nvars": null,
    "alpha": [
      0.001,
      0.01,
      0.1
    ],
    "time_limit": null,
    "gap_limit": null,
    "prune": true,
    "timeout": 600,
    "edgeConstraints": "edgeConstraints.json"
  }
]