Peter Grünwald's Thesis

The title of my thesis is

The Minimum Description Length Principle and Reasoning under Uncertainty

ILLC Dissertation Series DS 1998-03


Co-winner of the 1999 FoLLI Outstanding Dissertation Prize.

Click here for a short abstract; click here for a longer synopsis. Click here (or read on) for the thesis itself.

The thesis consists of three parts, which can be read (more or less) separately. Part I contains a basic introduction to the MDL Principle and to its central concept, the Stochastic Complexity. It can be read without knowledge of information theory. It also contains some new theoretical developments concerning MDL that focus on the following question:

under what circumstances is it safe or even advisable to use overly simplistic models for the data at hand?

Part II contains some applications of MDL and Bayesian Methods. Part III concerns work I have done in a somewhat different field: Nonmonotonic Temporal Reasoning, specifically Reasoning about Action and the Frame Problem.

Downloading the Thesis

You can obtain the thesis (in gzipped postscript) here. For uncompressed postscript (3 Megabytes) click here. The full thesis is 300 pages! If you only want to read parts of the thesis, you should obtain the introduction, the ending (bibliography, index etc.) and the part(s) you want by clicking below: For technical reasons the on-line version does not look as nice as the original version, which was professionally printed as a book in the beautiful Lucida font. You can order this book by contacting Marjan Veldhuisen of the Institute of Logic, Language and Computation. The price is FL 25,- for orders from Holland and FL 45,- (around 22 $) for orders from abroad.

Back to Peter's home page