Research interests
| My education and my scientific experience fall in the areas of Computer Science, Software Engineering, Operational Research, Artificial Intelligence, Machine Learning, Industrial Automation and Control Science. |  |
I am a regular reviewer for several international research journals, such as: European Journal of Operational Research, Computers & Operations Research, Swarm Intelligence, Transactions on Autonomous and Adaptive Systems, Journal of Technological Forecasting & Social Change.
Summarizing, my areas of expertise are:
- Numerical algorithms and mathematical programming methods
- Metaheuristics (Genetic Algorithms, Ant Colony Optimization, Particle Swarm Optimization, Memetic Algorithms, Variable Neighbourhood Search, Tabu Search, Simulated Annealing, Scatter Search,…) in Combinatorial and Continue Optimization: application to real problems
- Reactive Search and Intelligent Optimization
- Robust network optimization and sensitivity analysis: models and algorithms
- Multi-objective network modelling and algorithms
- Network problems, with special emphasis for Path-finding, Location, and Transportation problems
- Data Mining and Clustering methods
- Neural Networks and Cellular Neural Networks
- Stereoscopy and 3D vision systems
- Field Programmable Gate Arrays (FPGA)
In December 2008 I graduated with a Ph.D. in Operational Research at the
School of Information
Systems, Computing & Mathematics at Brunel
University. During my Ph.D. experience, reported in my Ph.D. thesis:
The development and application
of metaheuristics for problems in graph theory: a computational study,
some recent NP-hard combinatorial optimization problems formulated on graphs
and pertaining to different domains were investigated. These problems are:
- the Minimum Labelling Spanning Tree problem,
- the Minimum Labelling Steiner Tree problem,
- the Minimum Quartet Tree Cost problem.
The first problem investigated, the minimum labelling spanning tree problem, is an NP-hard problem in which, given a graph with labelled (or coloured) edges, one seeks a spanning tree with the least number of labels (or colours). Such a model can represent many real-world problems in telecommunications networks, power networks, and multimodal transportation networks. For example, in telecommunications networks, there are many different types of communications media, such as optical fibre, coaxial cable, microwave, and telephone line. A communications node may communicate with different nodes by choosing different types of communications media. Given a set of communications network nodes, the problem is to find a spanning tree (a connected communications network) that uses as few communications types as possible. This spanning tree will reduce the construction cost and the complexity of the network. Because this problem is NP-hard, several metaheuristics were investigated. The proposed methods were compared to other algorithms recommended in the literature. Nonparametric statistical tests showed that the proposed heuristics outperform the other algorithms in the literature. Furthermore, a comparison with the results provided by an exact approach showed that these heuristics quickly obtain optimal or near-optimal solutions.
The second problem investigated, the minimum labelling Steiner tree problem, is an extension of the minimum labelling spanning tree problem to the case where only a subset of specified nodes, the basic nodes, need to be connected. This problem is NP-hard and, therefore, heuristics and approximate solution approaches with performance guarantees are of interest. Some metaheuristics for the problem have been proposed and, considering a wide range of problem instances, these metaheuristics outperformed the most popular heuristics already in the literature for the considered problem.
During the investigation of another NP-hard problem, the minimum quartet tree cost problem, large datasets were analysed and explored with data mining and intelligent data discovery techniques, and the extracted knowledge was modelled and formalised by means of particular families of graphs, called full unrooted binary trees, ternary trees, or boron trees. In particular, given a set of objects and their pairwise distances, it was required to determine an efficient visual representation of the data. The quartet paradigm, based on this NP-hard graph optimization problem, was used to compute a hierarchy of clusters of the objects. Several metaheuristic approaches were presented to approximate the optimal hierarchy. The performance of the proposed algorithms was tested through extensive computational experiments, and it was shown that they are able to obtain high quality solutions in short computational running times. In particular, in the experiments, data from different fields were considered in order to evaluate how the algorithms are influenced by the nature of the objects. Some examples from nature concerning a study in genomics with DNA sequences of different placental mammalian species have been considered, along with other examples with real geographic distances between famous cities, and other data obtained by mining of the World Wide Web through an automatic web information extraction method. It has been showed, for example, how to build ontologies of famous historical persons or popular musical artists by means of data extracted from the World Wide Web.
In April 2005 I received the M.Sc. degree in Computer Engineering with specialization in Industrial
Automation and Control Technologies from the University of Catania.
During my experimental M.Sc. thesis I worked for nine months in the Department of Systems, Electrical and Electronic Engineering at
University of Catania, with the cooperation of the Automation group of the multinational STMicroelectronics,
on the project titled: Project of a digital interface for the ACE16K chips based on Field Programmable Gate Arrays. The project
regarded the realization of a powerful board exploiting the full performances of a 128*128 Cellular Nonlinear
Network Analog Device (ACE16K) chip and its core in order to study the emergent chaotic dynamics.
Its signal processing capabilities needed a powerful interface to be deeply exploited, so the projected board
used the Field Programmable Gate Arrays (FPGA) technology to have the complete control of the ACE16K
pinout. The hardware system was mainly based on two chips: the ACE16K and the FPGA, which interfaced
the 128x128 nonlinear cells of the ACE16K with a PC, where a high-level program permitted interaction with
the system. The conceived platform interfaced with a computing environment in order to give a user-friendly
interface jointly with high level performances to the research users.
