wearepal / baynet Goto Github PK

you didn't ask for this, but I wanted to try out mutmut and the tests on this repo run the quickest out of everything.

The great news is that there aren't that many mutants (as opposed to EthicML that has like 50% mutants survived).

But I'm sharing the report in case you get really bored one day and decide that you want to squash some mutants

html.zip

Describe the bug
On the latest Mac OS 11, python-igraph 0.7.1.post7 will not compile an I've spent too long trying to find a compiler version that will make it work.

To Reproduce
Latest MacOS, pip install -e .

Expected behavior
Should just work

Screenshots
lots and lots of compiler errors

System Settings (please complete the following information):

• OS: MacOS
• Version 11.0.1

Additional context
python-igraph 0.8.3 (the latest version) installs just fine.
The conda-forge python-igraph 0.7.1.post7 also doesn't work for me.
The tests don't pass with 0.8.3 because some super constructors fail.
Obviously this affects Intervene too

As a: User
Given: A DAG and a dataset
When: I run estimate_parameters with option dfe specified
Then: The parameters of the BN are learnt using DFE.

Add parameter learning capability using MLE

Fix pickle with new serialisation method to make it compatible with the Multiprocessing library

Repurpose PGMPY's bif reading code, in order to load bayesian networks saved in the BIF format as a BayNet Object.

Is your feature request related to a problem? Please describe.
Not a problem - just a feature to be documented in manual.
ex.

priors:
  - variable: "dm_victim_race"
    level: "white"