PyUTAI

This repository contains a Python implementation for the potentials described in [1]. You can find the library in the folder pyuntai. The test files can be found in the test folder.

Python API

This module contains the different implementations of potentials that are to be implemented.

Initially it will contains the class Tree, that is a wrapper class for a tree root node.

Typical usage example:

# data is read from a numpy ndarray object
  data = np.array(get_data())
  variables = ['A', 'B', 'C']
  cardinality= {'A':4, 'B':3, 'C':3}

  tree = Tree.from_array(data, variables, cardinality)

  # We can perform most of the operations over tree. For example:
  tree.prune()
  tree.access({'C':1, 'B':2})

Test

In the test folder we have scripts that implement test classes based on unittest. To run all unittest use:

python -m test

from this directory. If you only want to execute a particular test module, then run:

python -m test.my_module_name

Experimentals Results

Work in progress.

120c955e2725286652e4d17a6ec92cfdee3dea0f

References

[1] Gómez‐Olmedo, Manuel, et al. "Value‐based potentials: Exploiting quantitative information regularity patterns in probabilistic graphical models." International Journal of Intelligent Systems 36.11 (2021): 6913-6943.