Research Areas

  • Dimensionality reduction and data visualization  
  • The goal of this emergent research area is to link the field of dimensionality reduction (DR) with that of information visualization (IV), in order to harness the special properties of the latter within DR frameworks. In particular, the properties of controllability and interactivity are of interest, which should make the DR outcomes significantly more understandable and tractable for the (no-necessarily-expert) user. These two properties allow the user to have freedom to select the best way for representing data.

  • Intelligence for embedded systems.  
  • This reseach line's main goal is to implement machine learning techniques into Embedded Systems within limited-resources frameworks. Therefore, it explores not only low-computation-cost machine learning techniques but preprocessing approaches (such as data cleansing and dimensionality reduction algorithms) aimed at reducing the computacional load and the usage of microprocessor's RAM memory. Given the versatility of sensor networks, the embedded systems may acquire data of interest from several fields: bio-signals, agriculture, and education, among others.

  • Case based reasoning (CBR) for medical applications.  
  • CBR in the health science are oriented to the case representation where is necessary synthesizing adequate features for CBR, reducing the number of features in highly dimensional data and one important focus will be how case based reasoning can associate probabilities and statistics with its results and taking into account the concurrence of several ailments.

  • Indentification of cardiac arrhythmias using unsupervised pattern recognition techniques.  
  • This research line's aim is to explore the benefit of using non-supervised approaches for data representation and classification on the automatic identification of cardiac arrhythmias in ECG Holter recordings.

  • Dynamic data analysis based on non-supervised techniques.  
  • The analysis of dynamic or time-varying data has emerged as an issue of great interest taking increasingly an important place in scientific community, especially in automation, pattern recognition and machine learning. There exists a broad range of important applications such as video analysis, motion identification, segmentation of human motion and plane tracking, among others. Spectral matrix analysis is one of the approaches to address this issue. Spectral techniques, mainly those based on kernels, have proved to be a suitable tool in several aspects of interest in pattern recognition and machine learning even when data are time-varying, such as the estimation of the number of clusters, clustering and classification. Most of spectral clustering approaches have been designed for analyzing static data, discarding the temporal information, i.e. the evolutionary behaviour along time. Some works have been developed to deal with the time varying effect. Nonetheless, an approach able to accurately track and cluster time-varying data in real time applications remains an open issue. This research aims to design kernel-based dynamic spectral clustering using a primal-dual approach so as to carry out the grouping task involving the dynamic information, that is to say, the changes of data frames along time.