DTAI Projects
FWO: Probabilistic-Logical Learning
Period: 01-2005 → 12-2008
Subgroup: ml,krr
Type: project
Members:
- Hendrik Blockeel
- Maurice Bruynooghe
- Luc De Raedt
- Danny De Schreye
Machine learning and data mining refer to the process of analysing data with the goal of discovering regularities in them, which can be used to improve the performance on an automated task (machine learning) or offer new insights to the human user (data mining, knowledge discovery).
Classical methods require that data are transformed into a so-called standard format. This transformation is non-trivial and may yield information loss.
Relational learning methods (such as inductive logic programming) are morepowerful because they do not require such a transformation. On the other hand, there is no good theoretical framework to handle probabilistic knowledge in the relational / first order logic setting. Hence, the integration of probilistic information with relational information is currently an active research domain.
This project consists of a comparative study of probabilistic-logical representation languages; a comparative study of probabilistic relational/logical learning algorithms, where probabilistic first order logic is used as a reference point; and, as a result of this, novel contributions to a unifying framework for probabilistic-logical learning.
← return to the projects
