Dipl.-Inf. Thomas Geier

Thomas Geier
Dipl.-Inf. Thomas Geier
Research Associate
Ulm University
Institute of Artificial Intelligence
James-Franck-Ring
89081 Ulm
Baden-Württemberg
Germany
URL:

Thomas Geier

„Essentially, all models are wrong, but some are useful.“ - George E. P. Box

Thomas Geier has left Ulm University. This page is no longer maintained.

I finished my studies of computer science at the University of Würzburg in 2007 with a thesis in the area of automata theory, obtaining the degree of Diploma. After that, I was working mainly as an embedded programmer with some degression into electronics and the physics of thermal transfer printing.

Since 2009 I am a PhD student working within the Sonderforschungsbereich TR-62 (A Companion Technology). There I am associated with developing a knowledge base that mediates between declarative, deterministic models (ontologies, models for deterministic planning) and sub-symbolic, probabilistic model components emerging from machine learning.

My research focus lies with inference in probabilistic, graphical models - often involving a temporal aspect.

More generally I am interested in the following topics:

  • complexity theory
  • graphical models
  • SAT and CSP
  • statistics and causality
  • functional programming

Weitere Information

libexp

libexp is a library for writing computational experiments in Scala. It is hosted on Github.

UUMLN

UUMLN is a software for inference in Markov Logic Networks. It contains the code used in the paper "Approximate Online Inference for Dynamic Markov Logic Networks". It is written in Scala and the source code can be downloaded here.

Replication of Experiments for "Conditioned Belief Propagation Revisited"

Both the source code and the final data used for the paper and the technical report are available at recomputation.org inside a virtual machine. If you want to have the code and data without the virtual machine, either extract it or email me.