The aim of the Adaptive Computing group is the analysis and development of computational models and algorithms relying on data analytics and soft computing techniques for solving complex problems. The group’s activities are structured in three main research streams: machine learning and pattern recognition, personalized health care, and smart heterogeneous systems. The group has lead and participated in several national (FCT, FCG, QREN, P2020) and international projects (FP7, H2020, COST, EUREKA) and organized several international workshops and conferences.

In machine learning and pattern recognition current research topics include dynamic models for ensemble learning in data streams with drifts (social networks), graph mining algorithms (biological networks, financial applications). We devised novel probabilistic graphical models to account the distinct annotators’ expertise in crowdsourced data. Other strong topics are Gaussian Processes with active learning (e.g. transport systems) and multimodal learning with deep Boltzmann machines (image, text, video and audio).

In the healthcare, the group has developed and implemented innovative solutions for personalized non-invasive remote monitoring applications (pHealth), able to support the management of diseases and early diagnosis of critical events. Applications are in preventive medicine and in chronic disease management, namely cardiovascular and cardiopulmonary diseases, sleep and neurological disturbances (epilepsy).

In the context of smart systems, the focus includes resilient enhancement in networked supervisory systems and optimization over heterogeneous networks (fuzzy system, multi-agent based algorithms, data streaming analysis, and GIS based management of river and urban floods). Innovative solutions for remote and virtual laboratories have been developed to support research and training activities in collaborative environments for various targeted areas (platform development, testbed based applications and algorithms assessment).