sg.bern.xval_classifier: Cross-Validated Bernoulli Subgraph Classifier

In neurodata/subgraphing: Signal Subgraphs and Extensions

Description

sg.bern.xval_classifier Bernoulli Subgraph Classifier with Cross Validation to determine the optimal subgraph.

Usage

sg.bern.xval_classifier(samp, Y, nedge, coherent = FALSE, tstat = "fisher",
  xval = "loo", folds = NaN)

Arguments

samp	a list or array of graphs with arbitrary labelling. - if samp is a list, then it should have s elements of dimensions [n x n]. - if samp is an array, then it should be of dimensions [n x n x s].
Y	[s] the class labels.
nedges	[z] an array where each element is the number of edges to look for, arbitrarily breaking ties as necessary.
coherent=FALSE	if FALSE, estimate an incoherent subgraph, otherwise an integer indicating the number of vertices in the coherent subgraph.
tstat="fisher"	the test statistic to use. options are fisher's exact ("fisher") and chi-squared ("chisq").
xval="loo"	the cross-validation options to use. Options are "loo" (leave-one-out) and "kfold" (K-fold).
folds=NaN	the number of folds to do if xval is set to kfold.

Value

subgraph [n x n] an array indicating whether an edge is present or not present in the subgraph.

p [n x n x c] the probability per edge of being connected per class for c classes.

See Also

sg.bern.compute_graph_statistics

neurodata/subgraphing documentation built on May 21, 2019, 8:10 a.m.