Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/wannesm/PySDD

Python package for Sentential Decision Diagrams (SDD)
https://github.com/wannesm/PySDD

cython knowledge-compilation model-counting python sdd weighted-model-counting

Last synced: 4 months ago
JSON representation

Python package for Sentential Decision Diagrams (SDD)

Host: GitHub
URL: https://github.com/wannesm/PySDD
Owner: wannesm
License: other
Created: 2018-03-14T15:37:17.000Z (over 6 years ago)
Default Branch: master
Last Pushed: 2024-07-03T08:20:06.000Z (5 months ago)
Last Synced: 2024-07-29T06:52:53.751Z (4 months ago)
Topics: cython, knowledge-compilation, model-counting, python, sdd, weighted-model-counting
Language: Cython
Homepage:
Size: 1.62 MB
Stars: 53
Watchers: 7
Forks: 16
Open Issues: 5
Metadata Files:
- Readme: README.rst
- License: LICENSE

Awesome Lists containing this project

README

        =====

PySDD

=====

Python wrapper package to interactively use `Sententical Decision Diagrams (SDD) `_.

Full documentation available on http://pysdd.readthedocs.io.

------------

Dependencies

------------

* Python >=3.6

* Cython

Optional:

* cysignals

* numpy

Make sure to have the correct development tools installed:

* C compiler (see `Installing Cython `_)

* The Python development version that includes Python header files and static library (e.g. libpython3-dev, python-dev, ...)

------------

Installation

------------

.. code-block:: shell

   $ pip install PySDD

--------------

Python package

--------------

The wrapper can be used as a Python package and allows for interactive use.

The following example builds an SDD for the formula ``a∧b ∨ b∧c ∨  c∧d``.

.. code-block:: python

    from pysdd.sdd import SddManager, Vtree, WmcManager

    vtree = Vtree(var_count=4, var_order=[2,1,4,3], vtree_type="balanced")

    sdd = SddManager.from_vtree(vtree)

    a, b, c, d = sdd.vars

    # Build SDD for formula

    formula = (a & b) | (b & c) | (c & d)

    # Model Counting

    wmc = formula.wmc(log_mode=False)

    print(f"Model Count: {wmc.propagate()}")

    wmc.set_literal_weight(a, 0.5)

    print(f"Weighted Model Count: {wmc.propagate()}")

    # Visualize SDD and Vtree

    with open("output/sdd.dot", "w") as out:

        print(formula.dot(), file=out)

    with open("output/vtree.dot", "w") as out:

        print(vtree.dot(), file=out)

The SDD and Vtree are visualized using Graphviz DOT:

.. image:: https://people.cs.kuleuven.be/wannes.meert/pysdd/sdd.png

.. image:: https://people.cs.kuleuven.be/wannes.meert/pysdd/vtree.png

More examples are available in the ``examples`` directory.

An interactive Jupyter notebook is available in

`notebooks/examples.ipynb `_

----------------------

Command Line Interface

----------------------

A Python CLI application is installed if you use pip, ``pysdd``. Or it can be used

directly from the source directory where it is called ``pysdd-cli.py``.

This script mimicks the original sdd binary and adds additional features (e.g. weighted model counting)

.. code-block:: shell

    $ pysdd -h

    $ ./pysdd-cli.py -h

    usage: pysdd-cli.py [-h] [-c FILE | -d FILE | -s FILE] [-v FILE] [-W FILE]

                    [-V FILE] [-R FILE] [-S FILE] [-m] [-t TYPE] [-r K] [-q]

                    [-p] [--log_mode]

    Sentential Decision Diagram, Compiler

    optional arguments:

      -h, --help  show this help message and exit

      -c FILE     set input CNF file

      -d FILE     set input DNF file

      -s FILE     set input SDD file

      -v FILE     set input VTREE file

      -W FILE     set output VTREE file

      -V FILE     set output VTREE (dot) file

      -R FILE     set output SDD file

      -S FILE     set output SDD (dot) file

      -m          minimize the cardinality of compiled sdd

      -t TYPE     set initial vtree type (left/right/vertical/balanced/random)

      -r K        if K>0: invoke vtree search every K clauses. If K=0: disable

                  vtree search. By default (no -r option), dynamic vtree search is

                  enabled

      -q          perform post-compilation vtree search

      -p          verbose output

      --log_mode  weights in log

    Weighted Model Counting is performed if the NNF file containts a line

    formatted as follows: "c weights PW_1 NW_1 ... PW_n NW_n".

-----------------

Memory management

-----------------

Python's memory management is not used for the internal datastructures.

Use the SDD library's garbage collection commands (e.g. ref, deref) to

perform memory management.

-----------------------

Compilation from source

-----------------------

.. code-block:: shell

   $ pip install git+https://github.com/wannesm/PySDD.git#egg=PySDD

The repository should contain all the required files and libraries (unless

you use Windows). If you want to compile from source, note that some c-source

files from the SDD package have been updated to work with this wrapper and are

included in this repository. Do not overwrite these new files with the original

files.

* Download the SDD package from http://reasoning.cs.ucla.edu/sdd/.

* Install the SDD package in the PySDD package in directories

  ``pysdd/lib/sdd-2.0`` and ``pysdd/lib/sddlib-2.0`` without overwriting

  the already available files.

* Run ``python3 setup.py build_ext --inplace`` or ``make build`` to compile the

  library in the current directory. If you want to install the library such

  that the library is available for your local installation or in your virtual

  environment, use ``python3 setup.py install``.

For some Linux platforms, it might be necessary to recompile the libsdd-2.0 code with

the gcc option ``-fPIC`` and replace the ``pysdd/lib/sdd-2.0/lib/Linux/libsdd.a``

library with your newly compiled version.

The Windows platform is not supported. There is some initial support but we cannot

offer guarantees or detailed instructions (but are happy to accept pull requests).

----------

References

----------

This package is inspired by the SDD wrapper used in the probabilistic

programming language `ProbLog `_.

References:

* Wannes Meert & Arthur Choi, PySDD,

  in `Recent Trends in Knowledge Compilation

  `_,

  Report from Dagstuhl Seminar 17381, Sep 2017.

  Eds. A. Darwiche, P. Marquis, D. Suciu, S. Szeider.

Other languages:

* C: http://reasoning.cs.ucla.edu/sdd/

* Java: https://github.com/jessa/JSDD

-------

Contact

-------

* Wannes Meert, KU Leuven, https://people.cs.kuleuven.be/wannes.meert

* Arthur Choi, UCLA, http://web.cs.ucla.edu/~aychoi/

-------

License

-------

Python SDD wrapper:

Copyright 2017-2018, KU Leuven and Regents of the University of California.

Licensed under the Apache License, Version 2.0.

SDD package:

Copyright 2013-2018, Regents of the University of California

Licensed under the Apache License, Version 2.0.