|Dino Pedreschi, Franco Turini e Fosca Giannotti, Pisa KDD Laboratory, Universita' di Pisa e ISTI-CNR||Second semester|
The technologies of mobile communications and ubiquitous computing pervade our society, and wireless networks sense the movement of people and vehicles, generating large volumes of mobility data. This is a scenario of great opportunities and risks: on one side, mining this data can produce useful knowledge, supporting sustainable mobility and intelligent transportation systems; on the other side, individual privacy is at risk, as the mobility data contain sensitive personal information. A new multidisciplinary research area is emerging at this crossroads of mobility, data mining, and privacy.
This course precisely addresses the problem of privacy and anonymity of location and mobility data, on the basis of the experience gathered in a large-scale European research initiative, the GeoPKDD project â€“ Geographic Privacy-aware Knowledge Discovery and Delivery http://www.geopkdd.eu
â€“ which is coordinated by the instructors.
In its introductory part, the course will introduce:
- the advocated scenario of location-aware and movement-aware data, services and analytical opportunities;
- the privacy and anonymity threats and the privacy-preserving and anonymity- preserving techniques developed in the general case of relational data for supporting safe data publishing and data mining.
On this basis, the course shall focus on the issue of privacy and anonymity in location-aware and movement-aware data, presenting an original account of the newly emerging and active research areas of:
- privacy and anonymity in location-based services: models of location-privacy that support anonymity of mobile users during real-time service;
- privacy and anonymity in mobility data analysis: models of privacy and anonymity threats for mobility data publishing and mobility data mining, and related privacy-preserving techniques for secure mobility data publishing and secure mobility data mining.
|Ercan Kuruoglu, ISTI-CNR, Pisa||May 5-30|
Bayesian data analysis methods have become increasingly important in many fields such as economics, bioinformatics, computer science in the last decade. In particular, the Bayesian approach has received an ever increasing momentum in the field of signal processing, and classical statistical signal processing has been revolutionised into Bayesian signal processing. This change has been reflected immediately to applications such as image segmentation, audio restoration and machine vision with great success. This course aims to equip interested graduate students with the necessary theoretical background and the technical tools and also to present them various applications where Bayesian data analysis had and would have important potentials. Latest developments in the Bayesian field such as particle filters will also be presented which are hoped to lead to new research ideas in the classroom.
- Introduction to Bayesian Estimation Theory
- Bayes theorem,
- Classical versus Bayesian interpretation of probability,
- Bayesian philosophy
- Posterior distribution
- Point estimation based on posterior
- Prior distribution
- Non-informative priors
- Jeffrey's prior
- Conjugate priors
- Single parameter models, multiparameter models
- Why sampling?
- Reverse sampling
- Rejection sampling
- Importance sampling
- Monte Carlo sampling
- A review of Markov chain theory
- Markov processes
- Stationary distribution
- Markov chain Monte Carlo
- Relation with statistical mechanics
- Metropolis method
- Gibbs sampling
- Relation with simulated annealing
- Reversible jump MCMC
- Advanced topics: convergence, multiple MCMC, etc
- Sequential Monte Carlo
- Filtering non-stationary signals (Kalman filter and its extensions)
- Unscented Kalman filter
- Particle filter
- Applications (2 hours)
- Telecommunications , Image Processing, Bioinformatics , Computer vision , etc
|Paolo Ferragina e Fabrizio Luccio, Dipartimento di Informatica, Università di Pisa||March, April|
Modern information retrieval and data mining applications for the Web need to carefully cope with the specialties and size of Web structure and content in order to be efficient and scalable. In this course we will address these two issues discussing, on the one hand, the mathematical laws governing network growth and, on the other hand, the design of efficient algorithms that store and search huge amount of data.
In the first part, we will study the mechanisms according to which a network can grow under random and/or hand-directed rules, going from the classical "random graph process" to the "preferential linking" behavior of the Web. Mathematical analisys will be validated by experimental evidence. In particular, models of random and intentional networks attacks will be examined.
In the second part, we will address the new algorithmic challanges posed by the processing of large datasets and by the architectural features of the memory hierarchy of current computers. These issues force algorithmic designers and engineers to address simultaneously data compression (fitting more data in the faster/smaller memory levels) and cache-friendly data access (exploiting the accessing features of memory levels). We will survey basic and sophisticated techniques useful to cope with this new scenario.
|Antonio Cisternino, Vincenzo Gervasi, Dipartimento di Informatica, Università di Pisa||February 25 - March 7|
- Introduzione alle virtual machines JVM e CLR
- Specifica di una virtual machine
- Modello con Abstract State Machines della JVM
- Modello con semantica operazionale (Gordon e Syme)
- Lo standard CLI
- Type system
- Inheritance model
- Bounded parametric polymorphism
- Implementazione di una virtual machine: il CLR
- Rotor (JIT, Garbage Collector, Generics)
- Mono (differenze con Rotor)
- Struttura dei binari (dati e metadati)
- Compilare per una virtual machine
- Compilatori JVM e CLR
- ILX e i linguaggi funzionali
- I linguaggi dinamici
- Runtime Code Generation
- Manipolare il bytecode
- Generazione di codice al volo
- Annotated C# e Code Bricks
|Paolo Lisboa, John Moores University - Liverpool||February 7-14|
- 1-5. Generic principles of neural networks in medical applications
- Introduction to the concept of neural networks.
- Concept of self-organisation - SOM, overview of GTM; examples from Magnetic Resonance Spectroscopy (MRS).
- Automatic adaptive learning - ART model. Simple example from MRS.
- MLP with regularisation. Examples from medical classification - Cushing's data.
- Bayesian regularisation framework. Examples from medical classification â€“ Parkinson's.
- Basic structure of the SVM.
References [1, 2, 3, 4, 6, 7, 8, 9, 10, 13]
- 6-7. Survival analysis and prognostic modelling for oncology
- Examples from breast cancer prognosis and, hopefully, also applied to GIMEMA data.
References [11, 14]
- 38 Automatic rule extraction from data
- OSRE model.
- Examples from MRS, breast cancer and other medical data including some bioinformatics.
References [5, 12]
- 9-10 Clustering and visualisation for bioinformatics
- Optimisation of k-means models.
- Direct visualisation of clusters in 3-D.
- Example from Ferrara bioinformatics dataset
- Ripley, B.D. "Can Statistical Theory Help Us to Use Neural Networks Better?" Proc. Interface 97, 29th Symposium on the Interface: Computer Science and Statistics, 1997. http://www.stats.ox.ac.uk/~northrop/teaching/PAN/Interface97.pdf
- Lampinen, J. and Kostiainen, T."Overtraining and model selection with the self-organising map" Proc. IJCNN'99, Washington, DC, USA. http://citeseer.ist.psu.edu/420998.html
- Lisboa. P.J.G., Vellido, A. and Wong, H."Bias Reduction in Skewed Binary Classification with Bayesian Neural Networks" Neural Networks, 13, 407-410, 2000.
- Huang, Y., Lisboa, P.J.G. and El-Deredy, W."Tumour grading from Magnetic Resonance Spectroscopy: a comparison of feature extraction with variable selection" Statistics in Medicine, 22, 1, 147-164, 2003.
- Etchells, T.A. and Lisboa, P.J.G."Rule extraction from neural networks: a practical and efficient approach" IEEE Transactions on Neural Networks, 17 (2):374-384, 2006.
- Vellido, A. and Lisboa, P.J.G."Handling outliers in brain tumour MRS data analysis through robust topographic mapping" Journal of Computers in Biology and Medicine, 36 (10): 1049-1063, 2006.
- Barton, G.J., Lees, A., Lisboa, P.J.G. and Attfield, S."Visualisation of gait data with Kohonen self organising neural maps" Gait and Posture, 24, 1: 46-53, 2006.
- Lisboa, P.J.G. and Taktak, A.F.G."The use of artificial neural networks in decision support in cancer: A Systematic Review", Neural Networks, 19: 408-415, 2006.
- Barton, G.J., Lees, A., Lisboa, P.J.G. and Attfield, S."Gait Quality Assessment Using Self-Organising Artificial Neural Networks", Gait and Posture, 25, 3: 274-379, 2007.
- Fisher, A.C., El-Deredy, W., Hagan, R.P., Brown, M.C. and Lisboa, P.J.G."Removal of eye movement artefacts from single channel recordings of retinal evoked potentials using dynamical embedding and independent component analysis" Medical and Biological Engineering and Computing 45, 1:69-77, 2007.
- Taktak, A., Antolini, L., Aung, M.H., Boracchi, P., Campbell, I., Damato, B., Ifeachor, E.C, Lama, N., Lisboa, P.J.G, Setzkorn, C., Stalbovskaya, V. and Biganzoli, E.M."Double-blind evaluation and benchmarking of prognostic models in a multi-centre study" accepted for Computers in Medicine and Biology.
- Aung, M.S.H, Lisboa, P.J.G., Etchells, T.A., Testa, A.C., Van Calster, B., Van Huffel, S., Valentin, L. and Timmerman, D."Comparing Analytical Decision Support Models Through Boolean Rule Extraction: A Case Study of Ovarian Tumour Malignancy" accepted for Lecture Notes in Computer Science.
- Vellido, A., and Lisboa, P.J.G."Neural Networks and other Machine Leaning Methods in Cancer Research" accepted for Lecture Notes in Computer Science 4507-0996.
- Jarman, I.H., Etchells, T.A. and Lisboa, P.J.G."A Data-based Decision Support System for Breast Cancer Prognosis" accepted for Artificial Intelligence in Medicine.
- Lisboa, P.J.G., Ellis, I.O., Green, A.R., Ambrogi, F."Cluster-based visualization with scatter matrices" submitted to Pattern Recognition Letters.
|Roberto di Pietro, Dipartimento di Matematica, Universita' Roma Tre||May 5-16 (first lesson: May 5 16.00-18.00)|
5/5 Lu: 16-18 aula riunioni OVEST
6/5 Ma: 16-18 aula riunioni OVEST
7/5 Me: 9-11 aula riunioni OVEST
7/5 Me: 16-18 aula riunioni OVEST
8/5 Gi: 16-18 aula riunioni OVEST
12/5 Lu: 16-18 aula seminari EST
13/5 Ma: 16-19 aula seminari EST
14/5 Me: 16-19 aula seminari EST
15/5 Gi: 16-18 aula seminari EST
Mobile ad hoc networks (MANETs) and Wireless Sensor Networks (WSN) are able to provide fast and efficient network deployment capabilities in a wide variety of scenarios where a fixed networking infrastructure is not possible. Unfortunately, many proposed security solutions contain strong assumptions or they didnâ€™t well adapt to the characteristics of MANETs or WSNs. In this course we will review some of the security issues and solutions for MANET and WSN. Further, we will also address privacy issues in a two rising research areas, where the call for innovative solutions is demanding, that is Radio Frequency IDentifiers and vehicular ad hoc networking (VANET). The organization of the course is thought to raise also active participation of the attendees; at the end of each lecture, one hour is dedicated to discuss research issues/ideas that could foster cooperation among participants.
- Introduction to security issues in mobile, resource constrained devices. (2h)
- Confidentiality Issues in WSN (3h + 1 hour discussion)
- Replicated Nodes Detection in WSN (3h + 1 hour discussion)
- Authentication for resource constrained devices (3h + 1 hour discussion)
- Privacy issues in RFID and VANETS (2h + 1 hour discussion, 2h + 1 hour discussion)
|Frederico de Oliveira-Pinto, Universidade Indipendente, Lisboa, Portugal||June 23 - July 11|
- Recurrence and difference equations
- General solutions of discrete systems
- Equilibrium values
- Dynamical stability
- Examples of linear systems with exponential growth and known general solutions
- Analytical solutions of discrete linear versus continuos linear growth models
- Quadratic dynamical systems
- Discrete quadratic versus logistic systems
- Closed-form solutions for discrete quadratic and discrete cubic dynamical systems with "chaotic" behaviour
- Study of theis intrinsic instability
- Volterra-type predator-prey systems
- Military fighting models
23 June -- 4 July, everyday: 15-17 Sala Seminari EST
7 July -- 11 July, everyday: 15-17
- G. Fulford et al. - "Modelling with differential and difference equations", Cambridge University Prees, 1997.
- R.A.Holmgren - "A first course in Discrete Dynamical Systems", (2 Ed), Springer, 1996.
- F. Oliveira-Pinto and M. Adibpour - "Analitical solutions of one-dimensional Discrete Dynamical Systems with "chaotic" behaviour", Non-linear Dynamics, 1 (1990) p 121 - 129.
- F. Oliveira-Pinto and B.W. Conolly - "Applicable Mathematics of non-physical phenomena", Ellis Horwood, Chicherter, a Division of J. Wiley and Sons, 1982.
- J.T. Sandefur - "Discrete Dynamical Systems, Theory and Applications", Claredon Press, Oxford, 1990.
|Zhiming Liu, UNU-IIST, Macao||June 30 - July 11|
Model driven development is a most promising approach to separation of concerns in design of complex software systems. It employs a UML-like multi-view and multi-notational language for describing the models of the system at different stages in the development. It also supports component-based design and allows developers to design systems at a higher level of abstraction using models or specifications of components. They will be produced and integrated at a later implementation, assembly and deployment stage.
However, a major challenge in the practice of model-based software development is to ensure correctness and dependability of the software product. To deal with this challenge, we need a common semantics for the multi-view modeling approach, and a method for verification and reasoning about models of different views, correct refinement of models, and consistently relating the models of different views. Tool support for such a method needs to integrate sophisticated visual and editing tools for model construction, transformers for correctness preserving model transformations, checkers and provers for verification of properties of models. The method, techniques and tools have to be incorporated into existing software development processes and environment.
In this course, we introduce a method, called rCOS, that we recently developed for Refinement and Verification in Component-Based Model Driven Development. The outline of the content of the course is:
- Introduction: the state of the art in practical software engineering, the current status of formal methods, the motivation and theme of rCOS
- Unifying Theories of Programming (UTP) in a Nutshell: the semantic foundation of rCOS
- The theory of rCOS a). component-based modelling and refinement: interfaces, contracts, components, composition (connectors), coordination and glue. b). object-oriented modelling and refinement: classes, objects, references, polymorphism, subtyping, dynamic binding, oo refinement
- Development process and case study
- Tool support: UML profile for rCOS, model transformations and model verification and analysis
Schedule: Every day: 11-13 in Sala SEMINARI EST
Exam: select one of
- Group project: each group of two or three can work on their own identified project by using the rCOS method. It can be an object-oriented system, a component-based project, a project that needs both component-based and object-oriented modelling and design.
- Seminar: Relate rCOS to your own research interest
|Marcello Federico, Fondazione Bruno Kessler - Trento||May 7-19 (first lesson: May 7, 15.00 - 17.00)|
Machine Translation (MT) is one of the oldest and still far from being solved challenges taken on by computer science. The course will start by briefly reviewing the history, approaches, progress, and difficulties of MT. The central topic of the course will be the statistical MT approach introduced in the early 90's at IBM. In particular, the following topics will be covered: statistical framework of MT, word alignment models, log-linear models, training and search algorithms, speech and text translation, performance evaluation. Alternative MT approaches will be discusses, and current research trends in the field will be presented.