Desenvolupament de tècniques d'aprenentatge estadístic: xarxes neuronals, models gràfics, màquines kernel, tècniques de classificació, regressió, agrupament i visualització (manifold learning), aprenentatge actiu, semisupervisat, relacional, Bayesià, estructurat, i causal.
Development of automatic control algorithms for robotic systems and autonomous driving of vehicles in unstructured environments, using artificial intelligence techniques.
Large database analysis in which there are three characteristics that make them special: growth velocity, variety in the data classes and volume.
We are currently working on forms of knowledge representation that allow the supervision of the students tasks resolution, as well as the incorporation of affective support through the detection of emotions by means of typical techniques within the field of computer vision
Application of automatic learning techniques for problems with prediction, classification and recognition of patterns or trends.
Development of multi-agent simulation systems applied to social and healthcare systems, so that they can simulate not only the individual behaviour of each agent, but also the global behaviour of the system as an addition of individual behaviour.
Electronic toll collection systems and mobile vehicles. Satellite and telephony location systems. GPS and GSM applications. V2V, V2I and I2V communication systems. Bluetooth, RFID and WIFI detection systems. Cooperative systems. Dynamic navigation.
In addition to the experience acquired over the years by the members of the group in different areas of simulation, mainly in bioengineering and mechanical engineering, recent advances in machine learning have been assimilated by the group as another way of exploiting the results of its simulations. The group has applied these techniques to the development of pedestrain models applied to crowd simulation. It also allows a further step in transferring the results of the simulations to new medical and industrial sectors (i.e. predictors of ectopic foci, aneurysms, arrhythmia detectors, etc.).
Traffic monitoring, vehicle magnetic fingerprint identification systems, DAI systems, applications for route and/or time optimisation, travel times, O/D matrices, GIS applications, geographic data representation and virtual simulation. Emergency management plans.
Extraction of structured information and knowledge from the analysis of free texts and a priori unstructured information.
Recognition and machine learning techniques applied to linear and nonlinear models. Continuous Distance Learning. Heterogeneous and multi-dimensional data mining, with special emphasis on image analysis and recovery and the processing of audio and video signals.
Development of reinforcement learning models and dynamic programming for cost reduction, the improvement of important parameters and the increase of efficiency.
Development of product recommendation engines based on the characteristics of the customer and management of personalised promotions.
Multiple sources of medical images are routinely used in clinical practise (MRI, TAC, PET, ECO...) Methods are needed to isolate the organs, analyse their deviation from normality and matching them with other cases or with explorations of the same patient to help doctors in the diagnosis of diseases.
Simulation models for animation of complex behaviours in interactive graphics applications, including physical behaviours (fluids, granular systems, etc.) and behaviours based on artificial intelligence (reinforcement learning, etc.).
We model dynamic processes using spatial point processes and random closed sets. Descriptors for univariate and bivariate processes with an explicit statistical analysis of the time and tests for hypothesis testing are proposed. Software for simulation and analysis is developed. More info and applications to cell biology in https://www.uv.es/tracs.
Aplicacions en tractament de senyals i imatges de teledetecció: estimació de paràmetres biofísics i variables de fluxos, inversió de models, segmentació d'imatges, detecció de canvis i anomalies, fusió d'imatges i multiresolució, restauració, causalitat i atribució, rànquing.
The purpose of a content based image retrieval system is to retrieve all the images relevant to a user query. Our main goals are the study of new low level descriptors and relevant feedback procedures that are able to deal with the semantic gap, currently using also fusion with textual information.
