Alchemy 2 - Inference and Learning in Markov Logic

This is a fork of the code located at https://code.google.com/p/alchemy-2/.

Features added in this fork:

• Inference using Gurobi ILP solver.

Dependencies

• g++ 4.1.2
• Bison 2.3
• Flex 2.5.4
• Perl 5.8.8

You can install these using Homebrew on Mac or apt-get on Ubuntu.

On OSX

You just need to install gcc and perl. Bison and Flex must be present already.

$ brew install gcc perl518

On Ubuntu

Do not install Bison through apt-get as it will install the latest version (v3.0.2). Instead, build it manually.

$ wget http://ftp.gnu.org/gnu/bison/bison-2.7.tar.gz
$ tar xfz bison-2.7.tar.gz
$ cd bison-2.7
$ ./configure
$ make
$ sudo make install

Flex and Perl can be installed through apt-get.

$ sudo apt-get install flex perl

Gurobi

Gurobi is a commercial Integer Linear Program solver which demonstrates very high computational speeds.

Installation: OSX instructions, Linux instructions

You also need to obtain a license (Academic or otherwise) to be able to use it.

Building

On OSX

$ cd path/to/alchemy-2/
$ mkdir bin/obj
$ cd src
$ cp makefile_osx makefile

On Linux

$ cd path/to/alchemy-2/
$ mkdir bin/obj
$ cd src
$ cp makefile_linux makefile

Modify the makefile to set the Gurobi directory if it is not installed in the default places.

$ make all -j4

Usage

To perform inference with Gurobi, pass the flags "-a -ilp" with the command line arguments. For example,

$ ./bin/infer -i exdata/voting-gen.mln -e exdata/voting-test.db -r exdata/voting-test-ilp.out -q Democrat -a -ilp

You will need to execute learnwts.sh to generate voting-gen.mln

References

This piece of code was written as part of my following paper pdf

@inproceedings{nagaraja2014feedback,
  title={Feedback Loop Between High Level Semantics and Low Level Vision},
  author={Nagaraja, Varun K and Morariu, Vlad I and Davis, Larry S},
  booktitle={Computer Vision-ECCV 2014 Workshops},
  year={2014}
}

The conversion of MAP inference in MLNs to an Integer Linear Program is based on

@inproceedings{noessner2013rockit,
  title={RockIt: Exploiting Parallelism and Symmetry for MAP Inference in Statistical Relational Models.},
  author={Noessner, Jan and Niepert, Mathias and Stuckenschmidt, Heiner},
  booktitle={AAAI Workshop: Statistical Relational Artificial Intelligence.}
  year={2013}
}

License

By using Alchemy, you agree to accept the license agreement in license.txt

src/ contains source code and a makefile. doc/ contains a change log, and a manual in PDF, PostScript and html formats. exdata/ contains a simple example of Alchemy input files. bin/ is used to contain compiled executables.

Please refer to the change log at http://alchemy.cs.washington.edu/ for the latest changes to Alchemy.