Markov logic. An interface layer for artificial intelligence. With contributions by Jesse Davis, Tuyen Huynh, Stanley Kok, Lilyana Mihalkova, Raymond J. Mooney, Aniruddh Nath, Hoifung Poon, Matthew Richardson, Parag Singla, Marc Sumner, and Jue Wang.

*(English)*Zbl 1202.68403
Synthesis Lectures on Artificial Intelligence and Machine Learning 7. San Rafael, CA: Morgan & Claypool Publishers (ISBN 978-1-59829-692-1/pbk; 978-1-598-29693-8/ebook). ix, 145 p. (2009).

Markov logic as a language that connects first-order logic and Markov networks is described, along with some algorithms and applications. Markov logic and Alchemy are combined for an interface layer for artificial intelligence.

Basic background on first-order logic and probabilistic graphical models is provided. The Markov logic representation is defined and it is shown how Markov logic unifies and builds on these concepts. Markov logic is compared to other representations that combine probability and logic.

State-of-the-art algorithms for reasoning and learning with Markov logic are presented.These algorithms build on standard methods for first-order logic or graphical models, including satisfiability, Markov chain Monte Carlo, and belief propagation for inference; and inductive logic programming and convex optimization for learning. Many of these methods have been combined and extended to handle additional challenges introduced by the rich Markov logic representation.

Several extensions of basic Markov logic representation are described. These extensions increase the power or applicability to particular problems. In particular, it is shown how Markov logic can be extended to continuous and infinite domains, combined with decision theory, and generalized to represent uncertain disjunctions and existential quantifiers. These extensions demonstrate that Markov logic can be adapted when necessary to explicitly support the features of new problems.

Applications of Markov logic to several real-world problems are explored, including collective classification, link prediction, link-based clustering, entity resolution, information extraction, social network analysis, and robot mapping. A brief introduction to Alchemy is also provided.

Basic background on first-order logic and probabilistic graphical models is provided. The Markov logic representation is defined and it is shown how Markov logic unifies and builds on these concepts. Markov logic is compared to other representations that combine probability and logic.

State-of-the-art algorithms for reasoning and learning with Markov logic are presented.These algorithms build on standard methods for first-order logic or graphical models, including satisfiability, Markov chain Monte Carlo, and belief propagation for inference; and inductive logic programming and convex optimization for learning. Many of these methods have been combined and extended to handle additional challenges introduced by the rich Markov logic representation.

Several extensions of basic Markov logic representation are described. These extensions increase the power or applicability to particular problems. In particular, it is shown how Markov logic can be extended to continuous and infinite domains, combined with decision theory, and generalized to represent uncertain disjunctions and existential quantifiers. These extensions demonstrate that Markov logic can be adapted when necessary to explicitly support the features of new problems.

Applications of Markov logic to several real-world problems are explored, including collective classification, link prediction, link-based clustering, entity resolution, information extraction, social network analysis, and robot mapping. A brief introduction to Alchemy is also provided.

Reviewer: Alex Nabebin (Moskva)

##### MSC:

68T27 | Logic in artificial intelligence |

68T05 | Learning and adaptive systems in artificial intelligence |

68-02 | Research exposition (monographs, survey articles) pertaining to computer science |

##### Keywords:

artificial intelligence; Markov logic; first-order logic; probabilistic logic; inductive logic programming; satisfiability; probabilistic graphical models; Markov networks; Markov chain Monte Carlo; machine learning; reasoning; belief propagation; collective classification; link prediction; link-based clustering; entity resolution; information extraction; social network analysis; robot mapping; Alchemy
PDF
BibTeX
XML
Cite

\textit{P. Domingos} and \textit{D. Lowd}, Markov logic. An interface layer for artificial intelligence. With contributions by Jesse Davis, Tuyen Huynh, Stanley Kok, Lilyana Mihalkova, Raymond J. Mooney, Aniruddh Nath, Hoifung Poon, Matthew Richardson, Parag Singla, Marc Sumner, and Jue Wang. San Rafael, CA: Morgan \& Claypool Publishers (2009; Zbl 1202.68403)

Full Text:
DOI