greedySearch: Greedy BAP search

In cnowzohour/greedyBAPs: Greedy BAP Learning Using Penalised Maximum Likelihood Score

Description

Greedy BAP search

Usage

greedySearch(cov.mat, n, mg.start = NULL, n.restarts = 0, mc.cores = 1,
  max.steps = Inf, max.iter.ricf = 10, neighbourhood.size = Inf,
  eps.conv = 1e-12, max.in.degree = Inf, dags.only = FALSE,
  direction = "both", verbose = FALSE)

Arguments

cov.mat	p-by-p covariance matrix
n	Sample size
mg.start	Starting graph(s). Must be either NULL, an adjacency matrix defining the mixed graph (if n.restarts is 0) or a list of adjacency matrices of length n.restarts+1. If NULL, a starting graph is uniformly sampled for each run.
n.restarts	Number of restarts
mc.cores	Parallelism (number of cores to be used). If set to 1, no parallelism is used. Should be at most n.restarts+1.
max.steps	Max number of greedy steps per run
max.iter.ricf	Max number of iterations in RICF
neighbourhood.size	If set to a finite number, only this many (randomly sampled) neighbours are considered at each greedy step
eps.conv	Minimal score improvement to continue search after each step
max.in.degree	Max in-degree of considered graphs
dags.only	Consider DAGs only?
direction	Can be "forward" (only adding or changing edges), "backward" (only removing or changing edges), or "both" (adding, removing, and changing edges)
verbose	Print extra log output?

Value

List containing: final.bap - Adjacency matrix of "winning" BAP final.score - Score of "winning" BAP all.baps - List of lists (one per greedy search run) of adjacency matrices of all visited BAPs all.scores - List of vector (one per greedy search run) of scores of all visited BAPs times - List of vectors (one per greedy search run) of cumulative running time for every step

cnowzohour/greedyBAPs documentation built on May 13, 2019, 1:36 a.m.