Last Day of Summer School

We'll talk about description logics, ontology matching and Markov Logic, and simple weight learning algorithms for ML. We'll also run some live experiments.

Here are the slides.

Fourth Day of Summer School

Markov logic, complexity of inference in graphical models, Sampling Methods, MaxWalkSAT, ILP

The slides are here.

Third Day of Summer School

First-order logic and undirected graphical models (aka Markov networks) are combined in Markov logic.

The updated slides are here.

Second Day of Summer School

We talked about probability theory, random variables, conditional independence, and Bayesian networks. We didn't quite get to undirected graphical models.

The slides are here.

Also, I mentioned Predictalot, Yahoo's combinatorial prediction market for the world cup. Here's the link. An interesting blog post about combinatorial prediction markets is here.

First Day of Summer School

As I promised I will also use this blog to upload my slides. Yesterday, I basically talked about my background, the motivation for knowing about and using Markov logic, and some basic propositional and first order logic.

Here are the slides of the first class

The other slides will be up at the end of each class at the latest.

I recommend participants of the class to download and install TheBeast. The link to the website is mentioned in the previous blog post.

Literature and Tools

Markov Logic is a probabilistic logic which combines the ideas
of Markov networks with those of first-order logic. It is one of the many languages that falls into the realm of statistical relational learning (SRL). SRL is concerned with models of domains that exhibit both uncertainty and relational structure.

Markov logic originates in the 2006 paper by Richardson and Domingos.

Some of the most successful applications can be found in the semantic web and natural language processing area:

• Natural Language Processing: Multilingual Semantic Role Labelling with Markov Logic, Ivan Meza Ruiz, Sebastian Riedel, CoNLL 2009
• Ontology refinement: Automatically Refining the Wikipedia Infobox Ontology, Fei Wu and Daniel S. Weld, WWW2008
• Ontology Matching: A Probabilistic-Logical Framework for Ontology Matching, Mathias Niepert, Christian Meilicke, Heiner Stuckenschmidt, AAAI 2010

Some open-source Markov Logic tools are available. We will mostly use TheBeast for in-class demos and experiments:

• Alchemy
• TheBeast

The Alchemy group at the University of Washington also maintains a list of Markov logic related publications.