dr.ir. T.J. (Tjalling) Tjalkens - Expertise
P.O. Box 513
5600 MB EINDHOVEN
Share / Bookmark
- D14340 - Measurement and control engineering
- D14310 - Telecommunication engineering
- D16300 - Theoretical computer science
- Machine learning
- Information theory - coding
- Data compression - source coding
Source coding. The research concentrates mainly on the field of (universal) noiseless source coding.
A report can be found regarding the universal variable-to-fixed length `modified Lawrence algorithm'. This paper discusses the achievable redundancy-rate (i.e. the best possible rate of convergence of the code rate to the source entropy) and presents an algorithm that achieves this optimal convergence.
Also a partial result on the redundancy-rate of the well-known Lempel-Ziv (1978) universal algorithm was presented at the 1990 IEEE Workshop on Information Theory. Here you can read our contribution to the 1990 Benelux Symposium on Information Theory, which contains a more elaborate discussion of the result.
The Context tree weighting algorithm.is a universal algorithm that has the property that for Tree sources of restricted depth the redundancy decreases as fast as possible. Tree sources are a large subset of Markov sources and they are especially useful in text modeling. Currently, the CTW algorithm achieves the best known compression on text data. (A table of results will follow soon).
Channel coding. Aplying enumerative methods to the design of `runlength-limited sequences' enabled us to describe Franaszek's `principal' states explicitely. Links CTW web page. Werkgemeenschap Informatie- en Communicatietheorie (WIC). Nederlands Elektronica- en Radio Genootschap (NERG). Euler Institute for Discrete Mathematics and its Applications (EIDMA). IEEE Information Theory Society.