gfn.gym.helpers.bayesian_structure.graph
========================================

.. py:module:: gfn.gym.helpers.bayesian_structure.graph


Functions
---------

.. autoapisummary::

   gfn.gym.helpers.bayesian_structure.graph.sample_erdos_renyi_graph
   gfn.gym.helpers.bayesian_structure.graph.sample_erdos_renyi_linear_gaussian


Module Contents
---------------

.. py:function:: sample_erdos_renyi_graph(num_nodes, rng, p = None, num_edges = None, node_names = None)

   Sample an Erdos-Renyi graph.

   :param num_nodes: Number of nodes in the graph.
   :param rng: Numpy random number generator.
   :param p: Probability of creating an edge.
   :param num_edges: Total number of edges (used to compute p if p is None).
   :param node_names: Optional list of node names.

   :returns:

             A PyTorch Geometric Data object representing the sampled graph with
                   an attribute 'node_names' mapping indices to node names.
   :rtype: Data

   :raises ValueError: If both p and num_edges are None.


.. py:function:: sample_erdos_renyi_linear_gaussian(num_nodes, rng, p = None, num_edges = None, node_names = None, loc_edges = 0.0, scale_edges = 1.0, obs_noise = 0.1)

   Sample a linear Gaussian Bayesian network based on an Erdos-Renyi graph.

   Creates graph structure using torch-geometric and assigns CPD factors for each node.
   Each CPD factor is constructed by sampling a parameter vector theta (with bias fixed to zero)
   based on the node's parent set determined from the graph structure.

   :param num_nodes: Number of nodes.
   :param rng: Random number generator for reproducibility.
   :param p: Probability of creating an edge.
   :param num_edges: Total number of edges (used to compute p if p is None).
   :param node_names: Optional list of node names.
   :param loc_edges: Mean value for edge parameters.
   :param scale_edges: Standard deviation for edge parameters.
   :param obs_noise: Observation noise for each node.

   :returns:

             A PyTorch Geometric Data object with additional attributes 'nodes' and 'cpds'.
                   'nodes' is a list of node names.
                   'cpds' contains a list of LinearGaussianCPD factors for each node.
   :rtype: Data