scratch of structure learning for discrete bn

[Roman223]

This pull request introduces several significant changes to the bamt package, focusing on enhancing the functionality and structure of the checker classes, graph classes, and node models. The most important changes include the creation of new checker classes, updates to the graph structure, and improvements to the node models.

Checker Classes Enhancements:

• bamt/checkers/base.py: Introduced the Checker class, which includes methods for validating arguments and initializing node types.
• bamt/checkers/network_checker.py: Added the NetworkChecker class, which extends Checker and includes methods for checking network restrictions, validating loads, and determining node types.
• bamt/checkers/node_checkers.py: Created RawNodeChecker and NodeChecker classes to handle raw and evolved node types, respectively. These classes include properties for checking node types and validation methods.

Graph Structure Updates:

• bamt/core/graph/dag.py: Modified the DirectedAcyclicGraph class to use pgmpy for directed acyclic graphs, added methods for topological sorting, and removed the old networkx graph references.
• bamt/core/graph/graph.py: Updated the Graph class to include nodes and edges attributes and added a __repr__ method for better representation.

Node Models Improvements:

• bamt/core/node_models/__init__.py: Corrected the import statements to use relative imports for better modularity.
• bamt/core/node_models/distribution.py: Added an abstract __repr__ method to the Distribution class for consistent representation across different distributions.
• bamt/core/node_models/empirical_distribution.py: Implemented the __repr__ method in the EmpiricalDistribution class and added an is_fitted method. [1] [2] [3] [4]

Node Classes Adjustments:

• bamt/core/nodes/__init__.py: Updated import statements to use relative imports for better modularity.
• bamt/core/nodes/child_nodes/child_node.py: Modified the ChildNode class to accept a name parameter in the constructor.
• bamt/core/nodes/child_nodes/conditional_discrete_node.py: Added a __repr__ method and placeholder methods for get_children, get_parents, and fit in the ConditionalDiscreteNode class.
• bamt/core/nodes/node.py: Updated the Node class to include name, disc_parents, cont_parents, and children attributes.
• bamt/core/nodes/root_nodes/continuous_node.py: Corrected the import statement for ContinuousDistribution.
• bamt/core/nodes/root_nodes/discrete_node.py: Modified the DiscreteNode class to accept a name parameter and updated the __repr__ method.
• bamt/core/nodes/root_nodes/root_node.py: Updated the RootNode class to accept a name parameter and added a __repr__ method.

Additional Changes:

• bamt/dag_optimizers/dag_optimizer.py: Updated the optimize method signature to accept data, scorer, and an optional formatter parameter.

[Copilot]

Pull Request Overview

This PR lays the groundwork for structure learning for discrete Bayesian networks by updating several constructors, method signatures, and import paths throughout the codebase. Key changes include:

• Changes to method signatures for node and optimizer classes to include a mandatory name parameter.
• Transition from str to repr in several classes to standardize object representations.
• Updates to import paths for improved package structure and relative imports.

Reviewed Changes

Copilot reviewed 40 out of 40 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
bamt/dag_optimizers/dag_optimizer.py	Updated the abstract optimize method signature to include additional parameters.
bamt/core/nodes/root_nodes/root_node.py	Modified constructor to require a name and changed str to repr.
bamt/core/nodes/root_nodes/discrete_node.py	Adjusted constructor signature, updated super().init usage, and modified string representation.
bamt/core/nodes/root_nodes/continuous_node.py	Revised import statements to relative imports.
bamt/core/nodes/node.py	Updated Node constructor to accept a name and initialized parent and children lists.
bamt/core/nodes/child_nodes/*	Modified constructors to include a name parameter and added repr methods.
bamt/core/node_models/*	Adjusted str to repr across distribution classes and fixed import paths.
bamt/core/graph/*	Updated the Graph and DAG classes with additional methods and revised node/edge handling.
bamt/checkers/*	Implemented or updated node and network checkers with new node type validations and error handling.

Comments suppressed due to low confidence (3)

bamt/dag_optimizers/dag_optimizer.py:9

• The optimize method signature was updated to include data, scorer, and an optional formatter. Ensure that all subclass implementations and corresponding documentation are updated accordingly.

def optimize(self, data, scorer, formatter=None):

bamt/core/nodes/root_nodes/root_node.py:19

• The method has been renamed from str to repr. Make sure that both documentation and any external calls to the string representation are updated to use repr.

def __repr__(self):

bamt/core/nodes/root_nodes/discrete_node.py:9

• Typo in 'DisscreteNode'; consider correcting it to 'DiscreteNode'.

Initialize the DisscreteNode with an optional EmpiricalDistribution.

[Copilot]

The check for discrete node appears incorrect by including the entire RawNodeType enum; consider using a specific member (e.g., RawNodeType.disc) instead of RawNodeType.

Suggested change

	return True if self.node_type in [RawNodeType, RawNodeType.disc_num] else False
	return True if self.node_type in [RawNodeType.disc, RawNodeType.disc_num] else False

[Copilot]

Using 'assert' with an exception instance is incorrect; replace it with 'raise TypeError("Wrong type of argument.")' to properly signal the error.

Suggested change

	assert TypeError("Wrong type of argument.")
	raise TypeError("Wrong type of argument.")