sumu.bnet module

Module for representation and basic functionalities for Bayesian networks.

class sumu.bnet.DiscreteBNet(nodes)

Bases: object

Represents a Bayesian network.

classmethod from_dag(dag, *, data=None, arity=2, ess=0.5, params='MP')
classmethod read_file(path_to_dsc_file)

Read and parse a .dsc file in the input path into a object of type DiscreteBNet.

Parameters:

load. (filepath path to the .dsc file to) –

Returns:

a fully specified Bayesian network.

Return type:

DiscreteBNet

sample(N=1)
class sumu.bnet.DiscreteNode(*, name=None, arity=None, cpt=None, parents=[])

Bases: object

sample(config=None)
class sumu.bnet.GaussianBNet(dag, *, data=None)

Bases: object

classmethod random(n, *, enb=4)
sample(N=1)
sample_params(data=None)
sumu.bnet.adj_mat_to_family_sequence(adj_mat, row_parents=False)
sumu.bnet.family_sequence_to_adj_mat(dag, row_parents=False)

Format a sequence of families representing a DAG into an adjacency matrix.

Parameters:

  • dag (iterable) – iterable like [(0, {}), (1, {2}), (2, {0, 1}), …] where first int is the node and the second item is a set of the parents.

  • row_parents (bool) – If true A[i,j] == 1 if \(i\) is parent of \(j\), otherwise a transpose.

Returns:

adjacency matrix

sumu.bnet.nodes_to_family_list(nodes)
sumu.bnet.partition(dag)
sumu.bnet.random_dag_with_expected_neighbourhood_size(n, *, enb=4)
sumu.bnet.topological_sort(dag)

Sort the nodes in a DAG in a topological order.

The algorithm is from [1].

sumu.bnet.transitive_closure(dag, R=None, mat=False)