rblip_asobs

ASOBS

Package	r.blip
Version	1.1
Language	R
Docs	here
Paper	Scanagatta et al.[1], Scanagatta et al.[2], Teyssier and Koller[3]
Graph type	DAG
Docker	bpimages/rblip:1.1
Module folder	rblip_asobs

Description

A score-based two-phase algorithm where the first phase aims to identify the possible parent sets, Scanagatta et al.[2]; Scanagatta et al.[1]. The second phase performs an optimisation on a modification of the space of node orders introduced in Teyssier and Koller[3], allowing edges from nodes of higher to lower order, provided that no cycles are introduced.

Example JSON

[
  {
    "id": "winasobs-bdeu",
    "scorermethod": "is",
    "solvermethod": "winasobs",
    "indeg": 80,
    "timeout": 180,
    "allocated": 80,
    "scorefunction": "bdeu",
    "alpha": [
      0.001,
      0.01,
      0.1
    ],
    "cores": 1,
    "verbose_level": 0
  }
]

[1] (1,2)

Mauro Scanagatta, Giorgio Corani, Cassio Polpo De Campos, and Marco Zaffalon. Approximate structure learning for large bayesian networks. Machine Learning, 107(8):1209–1227, 2018. URL: https://link.springer.com/article/10.1007/s10994-018-5701-9.

[2] (1,2)

Mauro Scanagatta, Cassio P de Campos, Giorgio Corani, and Marco Zaffalon. Learning bayesian networks with thousands of variables. Advances in neural information processing systems, 2015. URL: https://proceedings.neurips.cc/paper/2015/hash/2b38c2df6a49b97f706ec9148ce48d86-Abstract.html.

[3] (1,2)

Marc Teyssier and Daphne Koller. Ordering-based search: a simple and effective algorithm for learning bayesian networks. arXiv preprint arXiv:1207.1429, 2012. URL: https://arxiv.org/abs/1207.1429.