Alchemy 2 - Inference and Learning in Markov Logic

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

Description

Alchemy is a software package providing a series of algorithms for statistical relational learning and probabilistic logic inference, based on the Markov logic representation. Alchemy allows you to easily develop a wide range of AI applications, including:

• Collective classification
• Link prediction
• Entity resolution
• Social network modeling
• Information extraction

If you are not already familiar with Markov logic, we recommend that you first read the paper Unifying Logical and Statistical AI. If you want to understand how lifted inference algorithms operate, read the Probabilistic Theorem Proving paper.

Alchemy 2.0 includes the following algorithms:

• Discriminative weight learning (Voted Perceptron, Conjugate Gradient, and Newton's Method)
• Generative weight learning
• Structure learning
• propositional MAP/MPE inference (including memory efficient)
• propositional and lazy Probabilistic inference algorithms: MC-SAT, Gibbs Sampling and Simulated Tempering
• Lifted Belief propagation
• Support for native and linked-in functions
• Block inference and learning over variables with mutually exclusive and exhaustive values
• EM (to handle ground atoms with unknown truth values during learning)
• Specification of indivisible formulas (i.e. formulas that should not be broken up into separate clauses)
• Support of continuous features and domains
• Online inference
• Decision Theory
• Probabilistic theorem proving (lifted weighted model counting)
• Lifted importance sampling
• Lifted Gibbs sampling

More info at http://alchemy.cs.washington.edu/

Code

• 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.

Dependencies

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

You can install perl and gcc using Homebrew on Mac. Bison and Flex must be present already.

$ brew tap homebrew/versions
$ brew install gcc49 perl518

Build

Either git-clone or extract the downloaded archive in $PROJECT_HOME

• cd $PROJECT_HOME/src
• make depend
• make

Note: This fork of http://code.google.com/p/alchemy-2 has been updated to compile properly on a Mac following instructions from http://alchemy.cs.washington.edu/requirements.html

Usage

Structure learning

Learn the structure of a model given a training database consisting of ground atoms

learnstruct -i <input .mln file> -o <output .mln file> -t <training .db file>

Weight learning

Learn parameters of a model given a training database consisting of ground atoms

learnwts -i <input .mln file> -o <output .mln file> -t <training .db file>

Inference

Infer the probability or most likely state of query atoms given a test database consisting of evidence ground atoms

infer -i <input .mln file> -r <output file containing inference results> -e <evidence .db file> -q <query atoms (comma-separated with no space)>

Tutorial

Tutorial: https://alchemy.cs.washington.edu/tutorial/tutorial.html More: http://alchemy.cs.washington.edu/

License

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