Functional Loops over Time

[FelipeGiro]

Your checklist for this pull request

Please review the guidelines for contributing to this repository.

• Make sure you are requesting to pull a topic/feature/bugfix branch (right side). Don't request your master!
• Make sure you are making a pull request against the dev branch (left side). Also you should start your branch off our dev.
• Check the commit's or even all commits' message styles matches our requested structure.

Issue number(s) that this pull request fixes

• Fixes DynamicBayesianNetwork limitations #1340

List of changes to the codebase in this pull request

• substitute the creation of the Bayesian model creation to a build-in DynamicBayesianNetwork function get_constant_bn.
• As get_constant_bn already have the CPDs, removed add_cpds

NOTE:

Code is not fully functional. The line

pgmpy/pgmpy/inference/dbn_inference.py

Line 70 in c720120

start_markov_model.add_edges_from(combinations_slice_0[0])

add the edges in a different way, resulting in the following nodes:

NodeView(('A_0', 'A_1', 'B_1', 'C_0', 'B_0', 'C_1', <DynamicNode(A, 0) at 0x22298476f08>, <DynamicNode(B, 0) at 0x2229847f0c8>, <DynamicNode(C, 0) at 0x2229847f108>))

As you can see, the three last nodes were already added. I don't know the reason they are represented differently.

The question is: should we change how the node is represented/compared or should we come back to base.py and think another solution other than get_constant_bn

[ankurankan]

@FelipeGiro Sorry for the late reply. I am not sure what is the best way here. I remember that I had tried to change the get_constant_bn method to return a DynamicNode object for the nodes, but that caused a lot of trouble in integrating it with parameter learning code. And I couldn't find a nice way (without a lot of changes) to make it work with parameter learning and hence resorted to changing the node names with _. I am not sure if inference would work fine with the DynamicNode objects, but if it does, I would say we can add an additional argument to get_constant_bn to return either _ nodes or DynamicNode. Else, we can completely switch to the _ node names.