The mission IMATI "E. Magenes" is to encourage the development and diffusion of methods of computational mathematics and computer science in addressing problems and challenges for industrial innovation and society. Since its establishment, IMATI conducts research in areas related to Computer Science, Differential Modelling and Numerical Analysis and Mathematical Statistics.
RESEARCH ACTIVITIES IN COMPUTER SCIENCE
The research activities in Computer Science address a mix of topics in Mathematics and in Computer Science and apply the related skills and knowledge to produce original and innovative solutions to different applicative domains and problems. A distinguish aspect is the adopted methodology that combines modelling, formalisms and experimental approaches. These research activities address, especially for the Genova Unit, fundamental aspects of algorithms, semantics models and methodologies of computer science. Other topics are system knowledge for modelling and design of distributed and parallel systems; information systems data base and systems for information management.
These competences are mostly available at Genova Unit, where during the years two research groups in Computer Science have been active:
o Computing Architectures and High Performance Computing Group.
The group activity is reported on the official CNR web system, along the research line, or "commessa" named "Methodologies Algorithms and Applications of Cooperation Grid". This activity addresses the smart use of parallel and distributed, complex and heterogeneous systems in order to develop and to provide solutions, beyond the state of the art, for efficient and effective parallel algorithms and research e-infrastructure design and implementation for multidisciplinary scientific domains applications. Our research activity aims to design effective and efficient application software solutions able to cooperate with different levels of the architectural stack. We achieved important results, during the period from 2011 to 2014 in the following aspects: parallel algorithms for bioinformatics; scheduling strategies for Quality of Service (QoS) and energy cost reduction in distributed systems (mostly Cloud and Grid); design and implementation of e-infrastructures and Science Gateways for multidisciplinary research in Hydrology and Meteorology. Specific applications include cloud solutions for intelligent and sustainable mobility, business models analysis for SMEs that would like to provide cloud based solutions; computing platforms for High Impact Weather Events (HIWEs) study and monitoring.
The research group have three tenured members (one research director and two researchers) and attracts different young researchers and post- master thesis students and fellows. During the period from 2011 to 2014 the group supported by means of competitive research projects and grants the following people: 2 post-doc, one researcher ex art.23, and one senior "assegno di ricerca"; 2 PhD students (one from Russia) with multidisciplinary research projects; 4 "assegni di ricerca" for post master specialization and with participation to international research projects; 4 "tesi di Laurea" in computer engineering; 2 stages in the framework of international cooperation.
The budget dedicated to the support of young researchers, during the period from 2011 to 2014 exceeded 504.000 euros, completely obtained from competitive research projects and grants. Exit destinations of people include Research Foundations (CIMA and IIT) and private companies. The research group participated during the period to 5 international research projects (4 FP7 and one COST network) and to 11 national projects and contracts also as main contractor. The whole budget from projects exceeded 800.000 euors.
The scientific production includes 15 journal papers (5 class A following ANVUR rules). Members of the group were Guest Editors of two journal special issues for bioinformatics. Other five journal papers were accepted and will appear during 2015. Moreover, we add 4 book chapters and 18 papers in scientific conferences. We also organized different events including an international conference, 4 international satellite workshop with paper peer review and publication, 6 international workshops with abstract publications.
Further teaching activities includes courses in the University of Genova Computer Engineering Laurea Degree, PhD course about heterogeneous parallel systems and other PhD courses on multidisciplinary topics including DRIHM project summer school (http://www.isgtw.org/event/drihm-summer-school-2014-hydro-meteorology-e-science-madrid-spain).
The group is also involved in innovation and technology transfer activities. Researchers of the group are members Board of Directors of SIIT Technological District, and of the Scientific and Technical Committee of "Polo d'Innovazione" Transit and Sosia.
o Shape and Semantics Modelling: this research group, founded by Bianca Falcidieno at the beginning of the 80s, was one of the first dealing with Computer Graphics at CNR. Computer Graphics, which is legitimately part of Computer Science and Mathematics, is still a specific expertise of this group.
Its activities focus on all aspects related to the shape of 3D objects, ranging from methods for the representation and analysis of the geometry for the modeling, classification and recognition of 3D shapes, up to the formalization of knowledge and usage context of 3D objects. Furthermore, the research related to knowledge management also includes the study of methodologies for the management and for the semantic analysis of multi-dimensional data and of resources available on the web, such as data exposed via the Linked Data paradigm.
Fundamental research of Shape and Semantics Modelling group applies to several contexts, with excellence demonstrated by its outstanding scientific production and its participation in many international projects (FP7, H2020), in different fields: Cultural Heritage, Earth and Environment, Spatial Data Infrastructure, Industrial Design and Production, Medicine and Bioinformatics. In all these application contexts, the integration of geometric and semantic information plays a fundamental role in addressing the current massive digitalization of all kinds of information.
The group is nationally and internationally recognized for its expertise, and established important relations, mostly interdisciplinary, with major research institutions and universities with exchange of young scientists, undergraduates and graduate students. These include: Università di Genova (Facoltà di Matematica, Ingegneria e Informatica), Università di Bologna (Facoltà di Matematica), RWTH-AAchen/Germania (geometry processing e mesh repairing), University of Southern Australia (machine learning methods for the functionality analysis), Univ. of New York at Stony Brook/US (numerical methods for the analysis of 3D shapes), LIMOS and University of Clermont Ferrand/France (shape analysis), AMPT in Aix en Provence (shape modeling and analysis). The Group is also active in the Research and Innovation Pole POLITECMED (Polo Ligure delle TECnologie MEDicali) as a member of the Scientific and Directory boards.
Thanks to its outstanding fundraising ability and to its total re-investment in training and development of young researchers, the group currently involves 12 researchers and technologist (1 fixed-term contract), a first researcher and three research directors (1 associate). Over the years, the group also invested in training (2011 to 2014: 12 theses, 6 PhD, 7 PostDoc, 6 Research Fellows, 4 interns, 4 international schools organized).
The specific expertise of the Shape and Semantics Modelling Group is expressed in two research units ("commesse"), which are part of the DIITET Department.
The research unit: "Advanced techniques for 3D digital shapes analysis and synthesis" mainly deals with fundamental research related to the geometrical aspects of the study and development of methods for the creation, processing, analysis and classification of 3D and multi-dimensional shapes, obtained by acquisition or design of physical phenomena, of real and virtual objects.
- geometric and numerical methods for the representation, approximation and processing of 3D/ nD data;
- geometric and topological methods for the shape analysis
- geometric and topological methods for the comparison and similarity evaluation of 3D/nD shapes
- Multiresolution and multi-scale approaches for the manipulation of Big 3D Data.
The research unit: "Coding, elaboration and restitution of multidimensional media (MDM) knowledge" mainly focusses on the implementation of approaches to code and exploit the MDM knowledge, based on the realization of integrated tools for the formalization, processing, transmission, analysis, visualization, and restitution of the contained semantics.
In particular, it deals with:
- semantic descriptors for MDM coding;
- methods for the elaboration of the knowledge related to MDM and information resources (i.e. evaluation of similarity, granularity, inter-operability)
- methods for the exploration , analysis and annotation of MDM data considering the semantic aspects of information and their use in different applications ( historical business archives, environment , ... )
- tools for knowledge management in applications utilizing 3D digital content ( industrial design , medicine and bioinformatics , gaming and simulation, cultural heritage);
- methods for the quality assessment and consumption of resources exposed on the web according to the Linked Data paradigm;
To support its research activity, the group has a lab for laser scanner acquisition and for the printing of 3D objects.
In its long-lasting activity, the group has developed and maintains important infrastructures:
- Virtual Visualisation Service (VVS): e-infrastructure developed within EU FP7 INFRA Project VISIONAIR, based on the Digital Shape Workbench (DSW), which was created in the frame of the Network of Excellence AIM@SHAPE, coordinated by IMATI. This infrastructure supports the sharing and reuse of MDM models (images, videos, 3D models, cloud points...) and software for MDM analysis and processing. It is one of the most used repositories in computer graphics for algorithm benchmarking It offers various search modalities including a 3D search engine for 3D object retrieval based on shape similarity.
- Linked Thesaurus Framework for the Environment (LusTRE), developed in the frame of EU Project eENVplus to provide a solution to multicultural and multilingual problems in sharing of environmental data. Its relevant features are a Knowledge Infrastructure integrating via linked Data the best known and utilized Thesauri for the Environment and a set of Web Services for its exploitation. The framework is aimed at the annotation of information resources and at the support of data discovery. Some Thesauri were included in the Linked Open Data Cloud 2014, thus expanding its visibility. (http://linkeddata.ge.imati.cnr.it).
In the period 2011-15 the group has been involved in several projects: 6 EU , 7 regional, 2 MIUR premiali and 2 Factory of the Future Flagship (1 coordinated). Recently, the group has acquired 2 new H2020 projects, the first on the use of 3D geometric processing methods for the analysis of cultural heritage 3D objects (GRAVITATE), the second on mesh processing for 3D object representation for 3D printing (CaxMam) in collaboration with IMATI-Pavia.
RESEARCH ACTIVITIES IN THE FRAMEWORK OF STOCHASTIC MODELLING AND DATA ANALYSIS
The research activity belongs to disciplinary areas of statistical inference and probability.
Stochastic models for the interpretation of reality are selected, from within certain classes, by integrating knowledge on the nature of phenomena under investigation and an empirical analysis of observational data. The groups in Milan are specialized in dealing with space and time stochastic processes (such as point processes and Markov processes) and stochastic differential equations. To these, nonparametric methods are added, including multidimensional data classification methods, Bayesian nonparametric methods, and methods for the approximation of functions, which allow for less restrictive model classes. Furthermore, the institute has recently acquired new competencies in statistical design of experiments. The approach to statistical inference is often, but not exclusively, Bayesian, after the original scientific concept dating to the foundation, in 1980, of former IAMI in Milan.
Research groups in Milan are built around three main modelling and data analysis competencies: space and time stochastic processes, stochastic differential equations, nonparametric methods. Diffuse knowledge of traditional methodologies of statistical inference is also present in the groups.
During 2011-2014, thanks to funded research projects and to other forms of scientific exchange, the groups have attracted:
2 postdocs, who have contributed in full autonomy to the groups activities;
3 PhD students (one of whom from Germany), who have carried out part of their research at the IMATI;
4 research grant holders who participated in national research projects;
3 fixed term contracts for collaboration to research activity;
5 MSc students;
18 foreign visiting scholars.
Active research projects since 2011, divided by research group, are the following:
P1) Stochastic processes group
Joint Project CNR-Regione Lombardia: INTEGRATE - Technological Innovations for the Rational Management of the Built Environment (2013-2015), on the analysis of urban acoustic pollution data and on the comparison among different control policies of heating systems from the point of view of the energy efficiency, accounting for weather effects.
Progetto Congiunto CNR-Regione Lombardia: FIDEAS - Intelligent Factory for the Advanced and Sustainable De-manufacturing (2014-2015). A meta-model of the electrostatic separation of metal and non-metal particles, derived from the fragmentation printed circuit boards, has been obtained.
Integrated Technological Solutions for Zero Waste Recycling of Printed Circuit Boards (PCBs) (Zero Waste PCBs,2013-2014). Models for the characterization of the size distribution of particles, as they leave the shredder fragmenting PCBs, have been identified and estimated. In collaboration with the "nonparametric methods" group, the effectiveness of automatic classifiers of pictures of metal and non-metal samples taken by an hyperspectral camera has been analyzed.
Composite Nanofibres for Treatment of air and Water by an Industrial Conception of Electrospinning (Nano Twice,2013-2014). An algorithm for the detection of flaws in SEM images of polymer nanofibers has been developed, based on a dictionary and on the sparse representation of the images under evaluation.
EU Humanitarian Aid and Civil Protection Department (DG ECHO Unit A5), Project: Urban Prevention Strategies using Macroseismic and Fault sources (2012-2013) (UPStrat-MAFA - Num. 230301/2011/613486/SUB/A5). In collaboration with the "nonparametric methods" group a strategy for the classification of macroseismic fields based on some statistical indicators of the spatial distribution of seismic attenuation has been developed.
FP7-Capacities program: A web-based system for real-time Monitoring and Decision Making for Integrated Vineyard Management - MoDeM_IVM (FP7 reference number:262059; 2011-2012). The group contributed to the definition of a new Decision Support System by adapting and developing models for a few insect pests in vineyards and developing decision rules for vineyard management based on experts opinions.
Joint Project CNR-Regione Lombardia: New tools and technologies for energy efficiency and use of renewable sources in civil end-uses. Methods for the statistical characterization of distributed loads and co-generators have been developed.
P2) Stochatic differential equations group
Hospital Factory for Manufacturing Customized, Patient Specific 3D Anatomo-Functional Model and Prostheses (Fab@hospital, 2013-2014) ). It concerns enhanced methods and technologies required for implementing a prototyping service internal to the hospital for personalized prostheses and pre-operative planning.
Milano Defending Biodiversity: Solidarity and International Cooperation Grants 2009: Sustainable management strategies of agro-pastoral systems in Sub-Saharan Africa (2011-2012). A multi-trophic model, described by a system of differential equations, representing an agro-pastoral community, has been developed. This system is a constraint in an optimal control problem where the aim is to evaluate the optimal strategies of natural resources exploitation to maximize social welfare.
P3) Nonparametric methods group
Milano Defending Biodiversity 2008: Crop yield and climate change impacts on agriculture: adaptation strategies to desertification processes in the Mediterranean areas (2010-2011). The groups "stochastic processes" and "nonparametric methods" analyzed data about climate and annual yield production to evaluate the impact of climate change on cereals yield in Morocco under different climate change scenarios.
Outside research projects, the groups carry out theoretical and applied studies, which have produced results with an appreciable impact. These studies are described in the following, classified by research group.
1. Stochastic processes group. Stochastic processes are probabilistic models which can be used to describe the evolution of systems over time, space or on both. They are widely used to describe phenomena in environmental sciences, engineering, finance and health sciences, just to name a few fields.
Markov and point processes (especially Poisson processes) have been widely used as models in the area of reliability of repairable systems, where IMATI researchers are well known for their work on applications and on theoretical aspects, such as the recent new notion of signature. Dynamic and hidden Markov models are used to describe various aspects of industrial project management, with the goal of controlling final costs and delivery times. Stochastic models are used to study the aggregate behaviour of electrical loads and distributed generators.
The study of seismic sequences is another relevant area for which IMATI researchers have been known for years. Occurrence of strong earthquakes has been recently described by a multi-rupture model driven by a self-correcting process. Moreover studies have been carried out on the distribution of time between subsequent events. Other studies have led to identify clusters of space-time-magnitude sequences of earthquakes, and the development of hidden Markov models with observations from self-exciting models to identify anomalies in the seismicity rate of regions prone to slow earthquakes and seismic tremors.
Besides the two major areas of interest for IMATI (engineering and earthquakes), stochastic processes are considered in many other contexts, such as the analysis of terrorist events through self-exciting processes, maps of neonatal mortality through spatial processes and the evaluation of management strategies for habitat, climate change and extreme events, besides combination of experts' opinions in Bayesian networks.
The expertise of the IMATI researchers is such that they can use the same model, with minor changes, in different applied fields. Prime examples are the self-exciting processes and the hidden Markov models used in both reliability and earthquakes analysis; the former describe also terrorist activities and the latter are used in project management.
2. Stochastic differential equations. Stochastic differential equations (SDEs) are used to describe dynamical systems subject to random variations. They are widely applied in many fields when the noise affects the system. The IMATI group "stochastic differential equations" is mainly interested in the applications of SDEs in biology, finance and engineering.
SDEs are suitable to representing biological systems, in particular population dynamics where the state variables describe, for example, population densities. In this framework, an acarine predator-prey stochastic differential systems has been considered and methods of parameters estimation in SDEs, based on MCMC, have been developed. Following the same rationale, SDEs are applied to describe the dynamics of uninfected and infected cells during a Chlamydial infection.
In finance, SDEs are widely used to describe the dynamics of a security price. The group has developed methods for the evaluation of non-standard options where the dynamics of the underlying asset and of its stochastic volatility are represented through SDEs.
In the industrial field, the description of many processes is linked to differential systems, the coefficients of which are often unknown or known with uncertainty. In this context, the group has developed and experimentally validated estimation methods of the thermal conductivity in polymers, with the ability to assess the uncertainty associated with the estimates.
3. Nonparametric methods. This group has mainly worked with mixture models where the mixing distribution is a normalized completely random measure (e.g. the Dirichlet process or the normalized generalized gamma process), with a focus on computational issues. For a long time, efforts have been directed to a particular family of algorithms, called "conditional algorithms based on truncation" in the specialized literature.
Nonparametric mixture models have proved to be very powerful for dealing with clustering problems. The group has also have worked on this topic, focusing on data which are grouped into clusters with non-standard geometrically shapes. One promising application concerns a preliminary study on seismic events, to provide a criterion to associate these events with the seismogenic source that generated them.
Finally there are competencies in function approximation for signal reconstruction and multidimensional approximation, applied to the reconstruction of geophysical surfaces.
International visibility of the Milan section is granted by the scientific direction of the biennial BISP (Bayesian Inference in Stochastic Processes) and GDRR (Game ad Decision in Risk and Reliability) conference series. A new conference series, directed to young researchers, is the BAYSM (Bayesian Young Statisticians Meeting). Finally, a trademark activity is the ABS - Applied Bayesian Statistics Summer School, taking place every year since 2004, with the purpose of presenting recent developments in application of the Bayesian approach to statistical analysis, taught by scholars with a recognized international status.
The research activities described above have been carried out within the following Commesse/Moduli:
SP.P01.004: Development and application of statistical models
INT.P01.005: Simulation and assessment of environmental risk scenarios
ICT.P11.009: Stochastic system modelling
ICT.P10.014.001: Computational and statistical methods for the visualization of remotely sensed multidimensional data
TA.P02.030: Differential and stochastic modelling for the analysis of environmental phenomena
ME.P07.004.002: Statistical methods for biomedical problems
RESEARCH ACTIVITIES IN THE FRAMEWORK OF DIFFERENTIAL MODELING
The work carried out in the Pavia branch focuses mainly on the modeling, from the theoretical, numerical and computational point of view, of numerous problems from different application areas.
More specifically the research were focused on
Study of various aspects related to discretization methods, both classical and new generation, for computing the solutions of partial differential equations, with finite elements, both continuous and Discontinuous Galerkin type, finite volumes, wavelets, mimetic finite differences camouflage and virtual finite elements, NURBS (Non Uniform Rational B -Splines) :
o Theoretical properties (properties of the approximation spaces, stability ,convergence)
o Conforming and non conforming domain decomposition methods, and related substructuring preconditioners
o Stabilization methods
o Analysis and design of a posteriori error analysis and adaptive algorithms.
o Study of compatible approximations for diffusion problems with tensor coefficients and polyhedral meshes.
o Methods for the numerical solution of applied problems on parallel computing architectures.
Study of problems related to the mathematical analysis of partial differential equations of different types and from different areas of application
o Mathematical analysis methods for the treatment of nonlinear evolution problems with applications to phase change, to evolution problems of time-scaling invariant type, to electrocardiology.
o Modeling, simulation and analysis of the evolution of macroscopic thermomechanical special materials (in particular, the shape memory alloys).
o Analysis of systems of nonlinear conservation laws. This class of partial differential equations has important applications in areas such as fluid dynamics, traffic, both pedestrian and automotive, and others.
Development of innovative methods for the qualitative analysis and simulation
o Methods of qualitative analysis and simulation and their integration with classical methods for the identification and simulation of nonlinear dynamical systems. Applications in biology, particularly in the area of Systems and Synthetic Biology.
o Methods for the extraction and automatic interpretation of "features" from space / time data aimed at the qualitative interpretation of numeric fields, with applications to the study of electrocardiac activation maps.
o Learning from data methods, in particular, fuzzy logic systems.
Applications in different fields (electromagnetic, fluid dynamics, semiconductor devices, elasticity, material properties, fluid / structure interactions and electrocardiology) :
o Study of electromagnetism problems (solvers for the calculation of the eigenvalues , using discontinuous finite element and singular completion techniques). Compatible discretizations for time-harmonic problems
o Study and approximation of semiconductor nano-devices (double gate MOSFET).
o Study and approximation of hydrogeological models of groundwater flow and transport of pollutants, for the management and recovery of bodies of groundwater.
o Assessment of locking phenomena for finite elements in large displacements regime for materials with non linear inelastic behavior.
Methods of analysis and processing of signals and images: Atlases based image segmentation; Image registration; Applications in the medical field.
The research listed above have found further outlet in the following international projects:
- Project "Innovative compatible discretization techniques for Partial Differential Equations" (GeoPDES) , (2008-2013) funded by the European Research Council as part of a "Starting Independent Research Grant " . The research involves the study of compatible discretization schemes ie schemes that maintain in the discrete models the conservation properties of the underlying physical phenomenon, based on NURBS. The main advantages of the NURBS based methods are: the exact representation of the computational domain, the direct use of CAD products without the need for remeshing, and a significant improvement in the ratio between accuracy and computational cost.
- Project "Mathematics for Shape- Memory Technologies in Biomechanics " ( BioSMA ) (2008-2013) funded by the European Research Council as part of a "Starting Independent Research Grant " . Shape memory alloys (SMAs) are used for the realization of innovative devices that have a large impact on the development of biomedical applications (from arches for orthodontics to vascular stents) . Studies are in progress aimed at analyzing the macroscopic properties of these materials in order to make the numerical simulation more reliable and the design more effective not only in an experimental context but also for real biomechanical engineering problems.
- Research project KAU (2011-2013) The King Abdulaziz University in Saudi Arabia Jeddah- , entitled: "The analysis of mimetic finite difference methods and their applications to different problems such as flows in porous media (Darcy flow)"
- Research contract HUTCHINSON SA (2011-2012) entitled: "Using isogeometric techniques for simulating problems of nonlinear elasticity under large deformations ."
- Research contract Manifacture Française des Pneumatiques Michelin (2013-2015) entitled: "Evaluation of the potential use of the isogeometric approach for simulation of tire contact on dry or wet ground."
- Research contract TOTAL SA / HUTCHINSON SA (2014-2017) entitled: "Development, in an isogeometric environment, of numerical methods, algorithms and software for nonlinear problems of solid mechanics in large deformation regime."
The following national projects were also carried out :
- PRIN 2012- triennial project "Innovative methods in numerical differential modeling differential" entitled: "Development of numerical methods for the treatment of models based on partial differential equations". We aim at reducing the complexity of the current numerical models, in order to improve the impact of numerical modeling on the simulation, prediction and control of relevant phenomena in science, technology and society."
- Project Premiale 2012 "Mathtech - mathematics for society and technological innovation." The project is based on the use of mathematical methods in some of the priority action areas for the country and for the program Horizon 2020. Of particular interest, the issues related to health, intelligent transport, climate actions and factory of the future.
The researches described above were carried out in different CNR departments within the following: Commesse / Moduli:
o Development of expertise in modeling for the realization of large industrial projects (Dip. Production Systems)
o Modeling, analysis and simulation for shape memory devices (Dip. Production Systems)
o Advanced methods for modeling of multiscale and / or incomplete knowledge systems (Dip. ICT)
o Isogeometric Analysis for partial differential equations (Dip. ICT)
o Simulation and evaluation of environmental risks (Prog. Interdip. Safety)
o Module: Methods and algorithms for solving complex problems in high performance computing environments (Dip. ICT)
o Module: Analysis and synthesis of heterogeneous data for monitoring the degradation of Cultural Heritage (Dip. Cultural Heritage)
o Module: Advanced numerical methods for fluid mechanics (Dep. Of Earth and Environment)
o Module: Mathematical modeling of biological regulatory networks (Prog. Interdip. Machine Learning, Modeling and growing up)