Data Science Lecture Series: November 4, 2014
- 04 November
- 12:30 - 13:30
- 'Zwarte Doos', TU/e Campus
- Entry price:
- free after registration
The Context-Tree Weighting (CTW) method was proposed as a universal data-compression method in 1995 (Willems, Shtarkov, & Tjalkens, IEEE Trans. Inform. Theory, pp. 653-664, May).
It was demonstrated for this algorithm that it has an excellent compression performance, both in theory and in practice. In order to achieve good compression, the CTW method implicitly focusses on the tree-model that fits best to the observed data in the sense that this model achieves the smallest possible codeword length (minimum description length) for the data. A modified version of the CTW method outputs this best tree model.
Although the CTW method is based on context trees for processing tree models, there are generalizations of the CTW method based on more advanced context structures, that can handle more advanced model classes than the tree model class.
In the lecture the CTW method will be introduced, and some of its applications will be discussed.
If you would like to attend this lecture please register.