Machine learning
In Machine Learning, we are interested in:
-
- Unsupervised learning.
- Learning models expressed in symbolic representation: patterns or rules expressed in propositional logic or in first order logic.
- Different kinds of data: classical, relational, textual, temporal.
- Different applications: chemistry, environmental data, images, Geographic Information Systems, among others.
We have strong competence in the area of Relational Learning (also called Inductive Logic Programming --- ILP) that focuses on learning knowledge expressed in subsets of first-order logics. In this field, most of our studies regards supervised learning (classification) with positive and negative examples. We have also addressed problems in Statistical Relational Learning. There are other topics of our interest such as mining frequent patterns in relational databases. More generally, we are interested in learning models expressed by patterns or rules, dealing with qualitative or quantitative attributes and classical or relational data.
In the context of unsupervised learning (Clustering), for the past 10 years, members of our research team have been consistently conducting research in this field. In particular, we have proposed novel model of:
-
- Multi-view clustering: many data sets in real world are naturally comprised of different representations or views.
- Semi-supervised clustering: if some external knowledge is available, it can be employed to aid the unsupervised learning process.
- Multi-objective clustering: nowadays there are complex data that allow multiple interpretations, that is, several meaningful groupings for each object can be produced.
- Overlapping clustering: in some settings, like discovering communities in social networks, there is a natural overlap between the classes in the data.
- Constrained clustering modelled in the context of Constrained Programming.
Besides the theoretical issues associated with the development of these models, we are interested in the application of them to real world problems, particularly in the areas of text mining, information retrieval and bioinformatics.