graph_tool.inference.CliqueState

class graph_tool.inference.CliqueState(g, init_edges=False, init_max_cliques=True, init_list=None)

Bases: object

The state of a clique decomposition of a given graph.

Parameters:

gGraph

Graph to be modelled.

init_edgesbool (optional, default: False)

If True, the state will be initialized with edges being occupied.

init_max_cliquesbool (optional, default: True)

If True, the state will be initialized with the maximal cliques.

init_listdict (optional, default: {})

If given, this will give the initialization state. Keys are the clique nodes and values are the counts.

References

[young-hypergraph-2021]

Young, JG., Petri, G., Peixoto, T.P. “Hypergraph reconstruction from network data”, Commun Phys 4, 135 (2021), DOI: 10.1038/s42005-021-00637-w [sci-hub, @tor], arXiv: 2008.04948

Examples

>>> g = gt.collection.data["polbooks"]
>>> state = gt.CliqueState(g)
>>> state.mcmc_sweep(niter=10000)
(...)
>>> cliques = []
>>> for v in state.f.vertices():      # iterate through factor graph
...     if state.is_fac[v]:
...         continue                  # skip over factors
...     print(state.c[v], state.x[v]) # clique occupation
...     if state.x[v] > 0:
...         cliques.append(state.c[v])
...     if len(cliques) > 10:
...         break
array([0, 2, 4, 5], dtype=int32) 1
array([0, 1, 3, 5], dtype=int32) 1
array([0, 1, 5, 6], dtype=int32) 1
array([0, 4, 5, 6], dtype=int32) 0
array([2, 5, 7], dtype=int32) 1
array([ 4, 28], dtype=int32) 1
array([ 4,  6, 29], dtype=int32) 1
array([ 4, 30, 31], dtype=int32) 0
array([5, 6, 7], dtype=int32) 0
array([ 7, 71], dtype=int32) 1
array([ 7, 14, 58], dtype=int32) 1
array([ 7, 58, 85], dtype=int32) 0
array([ 7, 30, 58], dtype=int32) 1
array([ 6, 10, 12], dtype=int32) 0
array([ 6, 12, 18], dtype=int32) 1
array([ 8, 12, 13, 32], dtype=int32) 1

Methods

copy(**kwargs)	Copies the state.
entropy(**kwargs)	Get the description length, a.k.a.
get_nEd()	Get fraction of edge counts per clique size
mcmc_sweep([niter, beta])	Runs niter iterations of a Metropolis-Hastings MCMC, with inverse temperature beta, to sample clique decompositions.
reset_Ed()	Reset edge counts
reset_factors()	Reset factor constraints

copy(**kwargs)

Copies the state. The parameters override the state properties, and have the same meaning as in the constructor.

entropy(**kwargs)

Get the description length, a.k.a. the negative log-likelihood.

get_nEd()

Get fraction of edge counts per clique size

mcmc_sweep(niter=1, beta=1)

Runs niter iterations of a Metropolis-Hastings MCMC, with inverse temperature beta, to sample clique decompositions. This returns the change in description length and the number of moves.

reset_Ed()

Reset edge counts

reset_factors()

Reset factor constraints