probinet.models.dyncrep#
Class definition of DynCRep, the algorithm to perform inference in temporal networks [SCDB22].
Classes
|
Class definition of DynCRep, the algorithm to perform inference in temporal networks with reciprocity. |
- class probinet.models.dyncrep.DynCRep(err: float = 0.01, num_realizations: int = 1, max_iter: int = 1000, **kwargs)[source]#
Class definition of DynCRep, the algorithm to perform inference in temporal networks with reciprocity.
- check_params_to_load_data(**kwargs)[source]#
Check that the parameters to load the data are correct.
- compute_likelihood() float[source]#
Compute the likelihood of the models.
- Returns:
loglik – Log-likelihood of the models.
- Return type:
float
- enforce_constraintU(num: float, den: float) float[source]#
Enforce a constraint on the U matrix during the models fitting process. It uses the root finding algorithm to find the value of lambda that satisfies the constraint.
- Parameters:
num (float) – The numerator part of the constraint equation.
den (float) – The denominator part of the constraint equation.
- Returns:
lambda_i – The value of lambda that satisfies the constraint.
- Return type:
float
- fit(gdata: GraphData, T: int | None = None, mask: ndarray | None = None, K: int = 2, ag: float = 1.0, bg: float = 0.5, eta0: float = None, beta0: float = 0.25, flag_data_T: int = 0, temporal: bool = True, initialization: int = 0, assortative: bool = False, constrained: bool = False, constraintU: bool = False, fix_eta: bool = False, fix_beta: bool = False, fix_communities: bool = False, fix_w: bool = False, undirected: bool = False, out_inference: bool = True, out_folder: Path = PosixPath('outputs'), end_file: str | None = None, files: PathLike | None = None, rng: Generator | None = None, **_kwargs: Any) Tuple[ndarray, ndarray, ndarray, float, ndarray, float][source]#
Model directed networks by using a probabilistic generative models that assumes community parameters and reciprocity coefficient. The inference is performed via the EM algorithm.
- Parameters:
gdata – Graph adjacency tensor.
T – Number of time steps.
mask – Mask for selecting the held-out set in the adjacency tensor in case of cross-validation, by default None.
K – Number of communities, by default 2.
ag – Shape of gamma prior, by default 1.0.
bg – Rate of gamma prior, by default 0.5.
eta0 – Initial value of the reciprocity coefficient, by default None.
beta0 – Initial value of the beta parameter, by default 0.25.
flag_data_T – Flag to determine which log-likelihood function to use, by default 0.
temporal – Flag to determine if the function should behave in a temporal manner, by default True.
initialization – Initialization method for the models parameters, by default 0.
assortative – Flag to indicate if the graph is assortative, by default False.
constrained – Flag to indicate if the models is constrained, by default False.
constraintU – Flag to indicate if there is a constraint on U, by default False.
fix_eta – Flag to indicate if the eta parameter should be fixed, by default False.
fix_beta – Flag to indicate if the beta parameter should be fixed, by default False.
fix_communities – Flag to indicate if the communities should be fixed, by default False.
fix_w – Flag to indicate if the w parameter should be fixed, by default False.
undirected – Flag to indicate if the graph is undirected, by default False.
out_inference – Flag to indicate if inference results should be evaluation, by default True.
out_folder – Output folder for inference results, by default “outputs/”.
end_file – Suffix for the evaluation file, by default “_DynCRep”.
files – Path to the file for initialization, by default “”.
rng – Random number generator, by default None.
**kwargs – Additional parameters for the model.
- Returns:
u_f – Out-going membership matrix.
v_f – In-coming membership matrix.
w_f – Affinity tensor.
eta_f – Reciprocity coefficient.
beta_f – Beta parameter.
maxL – Maximum pseudo log-likelihood.