Ricerc@Sapienza

This paper deals with the opportunity of extracting useful information from medical data retrieved directly from a stereophotogrammetric system applied to gait analysis. A feature selection method to exhaustively evaluate all the possible combinations of the gait parameters is presented, in order to find the best subset able to classify among diseased and healthy subjects.

Computational intelligence and pattern recognition techniques are gaining more and more attention as the main computing tools in bioinformatics applications. This is due to the fact that biology by definition, deals with complex systems and that computational intelligence can be considered as an effective approach when facing the general problem of complex systems modelling. Moreover, most data available on shared databases are represented by sequences and graphs, thus demanding the definition of meaningful dissimilarity measures between patterns, which are often non-metric in nature.

Graphs are data structures able to efficiently describe real-world systems and, as such, have been extensively used in recent years by many branches of science, including machine learning engineering. However, the design of efficient graph-based pattern recognition systems is bottlenecked by the intrinsic problem of how to properly match two graphs. In this paper, we investigate a granular computing approach for the design of a general purpose graph-based classification system.

Workers healthcare gained a lot of attention recently as many countries are increasingly concerning about welfare. This paper faces the problem of predicting occupational disease risks by means of computational intelligence and pattern recognition techniques. Specifically, three different machine learning approaches are compared: the first one is based on the k-means algorithm, in charge to determine a set of meaningful labelled clusters as the final model. The latter two are based on fully supervised techniques, namely Support Vector Machines and K-Nearest Neighbours.

We propose in this paper a military application, which can be used in civil contexts as well, for solving inverse classification problems. Pattern recognition and decision support systems are typical tools through which inverse classification problems can be solved in order to achieve the desired goals. As standard classifiers do not work properly for inverse classification, which is an inherent ill-posed problem and therefore difficult to be inverted, we propose a new approach that exploits all the information associated with the decisions observed in the past.

This paper presents the electron and photon energy calibration obtained with the ATLAS detector using about 36 fb(-1) of LHC proton-proton collision data recorded at root s = 13 TeV in 2015 and 2016. The different calibration steps applied to the data and the optimization of the reconstruction of electron and photon energies are discussed. The absolute energy scale is set using a large sample of Z boson decays into electron-positron pairs.