Adaptive Computation Group

Our research concerns multimodal information processing for adaptive systems. Knowledge discovery in large heterogeneous data sets, with different scales, different uncertainty levels, sometimes incomplete, from different sources, with dynamic changes, to be used in prediction, feedback and decision is pursuited through computational intelligence techniques (neural networks, fuzzy systems, support vector machines, relevance machines, advanced signal processing, etc.,). Applications are in medical/clinical systems, industrial processes and internet distributed learning. The main research axes are:

  • eHealth for preventive and early diagnosis as well as supporting autonomous living through advanced monitoring and adaptive data analysis. Adaptive data analysis methods, including image analysis algorithms, are researched in order to develop non-invasive solutions for early diagnosis and prediction of life-critical or potentially life-critical events, such as heart diseases, sleep and neurological disturbances (epilepsy). Advanced monitoring research is focused on codec research for bio-signals and colour distortion models for accurate imaging in tele-medicine applications. The group participates in two FP7 projects one for heart deseases prevention other for epilepsy seizures prediction (this last as coordinator).


  • Incremental kernel algorithms specially designed for data analysis in large data sets with efficient and adaptive selection methods, with application to biological and physiological data analysis, namely in the context of a proposed project to FP7 about epilepsy seizures prediction.


  • Advanced kernel-based learning techniques applied to text classification in large data sets and internet scenarios, including active learning and hybrid strategies, powered by the increased availability of texts in digital form and the commanding need to organize them.


  • Spectral heterogeneous learning in structured, heterogeneous and distributed databases to provide computational tools for representing, integrating and modeling these databases, namely the massive amounts of biological data accessible in over a thousand of databases.


  • Metaheuristics to improve multidimensional scaling in high dimensional spaces (simulated annealing, genetic algorithms, tabu search) for classification of data with application to seizure prediction in epileptic patients (in the context of P7 Project) and supervision of industrial processes (applied to Sines Galp refinery).


  • Data analysis and dynamic systems modeling for safety control in critical industrial systems in order to support fault tolerance, control over the internet and monitoring of embedded control systems with application to a refinery (Galp), a paper machine (Soporcel) and safety control of railway systems (CP).


  • Performance control of wireless sensor networks in the framework of an international project submitted FP7 in September 2007.


  • Intelligent risk management to improve safety and reduce losses from natural railways hazards.