Research Topics

Ensemble methods

Development of algorithms for combining multiple predictive models in order to increase predictive performance compared to a single model. A specific topic we study in more detail, is ensemble pruning, the reduction of the size of an ensemble prior to combination in order to reduce complexity and increase predictive performance.

Learning from multi-label data

Development of algorithms that build accurate classification and ranking models from multi-label data and scale efficiently with respect to the number of examples and labels. More information is available here.

Text classification

We currently focus our research on textual data streams, such as e-mails and news feeds and the problem of concept drift which appears in these type of applications. In particular, we have developed an algorithm that considers the appearance of new predictive words over time. Due to its simplicity and effectiveness this algorithm is incorporated into an adaptive news reader named PersoNews. Furthermore, we have confronted the problem of recurring contexts in data streams. Finally, we have proposed a number of approaches for automated web service classification that consider textual and semantic features.

Reinforcement learning

We are concerned with the development of reinforcement learning methods for multiagent systems as well as the application of reinforcement learning in several domains like focused crawling and ensemble pruning. One particular topic of study within this framework is transfer-learning with the aim of speeding up the reinforcement learning process.

Knowledge discovery from biological data

Development of algorithms for building predictive models from biological data. Application of knowledge discovery approaches to biological data in order to gain new biological insights.
- Translation Initiation Site Prediction page.
- Polyadenylation Site Prediction page.

Distributed data mining

Development of distributed algorithms for knowledge discovery from physically distributed databases. Distributed construction of predictive models. Discovery of interesting knowledge with respect to similarities/differences of distributed databases. More information available here.

Learning for planning

Developing approaches for learning domain knowledge for selection among different planning methods based on data concerning the performance (speed and quality) of these methods. We have modelled the problem of planning method selection using various approaches and applied it for a) the automatic configuration of the planning parameters of planning systems and b) the selection among different planning systems.

Periodicity detection in temporal sequences

Developing algorithms for detecting multiple partial and approximate periodicities. Our approach is exploratory and follows a filter/refine paradigm. In the filter phase we introduce an autocorrelation-based algorithm that produces a set of candidate partial periodicities. The algorithm is extended to capture ap-proximate periodicities. In the refine phase we effectively prune invalid periodicities. We conducted a series of experiments with various real world data sets to test the performance and verify the quality of the results.