Research Topics

Research topics of ALDAPA are the following:

Comprehensible supervised learning algorithms

Comprehensible supervised learning algorithms

Within this line the researchers design and work with learning algorithms that can provide an explanation of the made classification in addition to an efficient classification. The research also focuses on problems where the number of examples of one of the classes is very imbalanced compared to the rest, class imbalance problem, what is a handicap for most of the machine learning paradigms.

Machine learning based behaviour modelling

Machine learning based behaviour modelling

The beginnings of this line where related to computer security were user and malware program behaviour was modelled. Lately, we are applying the experience in automatic behaviour modelling to the area of web interaction (Web Mining, Adaptive Web, People with restrictions, Sheltered social network) in diverse areas as for example, tourism and users with cognitive and/or physical disabilities. We are also applying machine learning based modelling techniques in the area of eGovernment and eServices.

Parallel computing

Within this line the group works with diverse parallel systems (clusters for distributed computing, shared memory multiprocessors, GPU, etc.). The group applies the high performance computing mainly in the context of simulation of physical phenomena (electronic interaction at molecular level, kinetic monte carlo, etc.)

Physiological computing

Physiological computing

The aim of this research line is the automatic treatment of physiological signals (ECG, EEG, sweating, etc.) using machine learning techniques to develop models to help people with special requirements (help for mobility, interaction in the environment, help for interaction with devices, etc.), detect anomalous emotional situations (stress, relax), etc.

History

The group started its activity in the 90s centring the research in pattern recognition and optimization. Diverse supervised and unsupervised learning paradigms –artificial neural networks, neighbourhood based classifies (kNN), multiple classifiers, clustering– were applied to diverse contexts: failure detection in electrical networks, automatic character recognition, optimization in merchandise distribution, fraud detection, customer fidelity, and computer security among others.