ESSIM & ECMI Summer School


European Summer School in Industrial Mathematics (ESSIM)

ECMI Summer School and Modelling Week

Università degli Studi di Milano, Italy, August 29- September 12, 2010


Contents

  1. 1 Preparatory School Course programs
    1. 1.1 MCMC Methods and Reliability of Model Predictions
    2. 1.2 Data-mining and statistical visualization
    3. 1.3 Introduction to Local and Global Optimization for Non Linear Programming
    4. 1.4 Level set methods for image reconstruction
  2. 2 Classes assignation
  3. 3 Modelling Week Projects Description
  4. 4 Schedule
  5. 5 Registration
  6. 6 Travel infos
  7. 7 Local Organizing Committee
The objective of the ESSIM Summer School is to confront Mathematics students with problems coming from Industry, where “Industry” is intended in a broad sense including Engineering, Material Sciences, Aeronautic, Biotechnologies, Optimization of industrial processes, Transports, Services and several other fields wherein industrial sector operates.
The Summer school is organized mainly for second cycle (master) students. Particularly brilliant students at the bachelor level and also a small amount of  PhD students, at the beginning of their PhD studies, can  be admitted.

The Summer School is part of the international master programmes developed in the framework of the Erasmus Curriculum Development project ECMIMIM.
The main activities consist in solving industrial problems suitable for mathematical modelling, performed in multi-national teams. These projects will be provided by instructors from universities which are partners in the European Consortium for Mathematics in Industry (ECMI).
The problems are chosen from the portfolio of industrial and commercial projects which the instructors or their institutions are currently involved in, thereby exposing the students to real world problems of production, technological or social interest.
 
The training will be conducted in a two-week School, where the first week will be dedicated to a preparatory school, consisting in 4 short courses on Mathematics useful for applications, each course will end with a seminar, held by an industrial speaker, showing the relevance of the mathematical subjects explained during the course to solve industrial problems and exposing the students to up to date European experiences of collaboration between Academy and Industry. Each student will have to follow 2 over 4 courses offered. All students will attend the industrial seminars. At the end of the first week the students will be evaluated with a written test.
 
The preparatory phase will be followed by a second week of team-based project work on specific problems coming from "Industry". The project work will mimic the problem solving approaches used in industry, with emphasis on mathematical modelling as a tool for innovations leading to improved products, manufacturing procedures, biotechnologies, optimized processes or services. Students collaboration and communication skills will also be trained.
The projects will be concluded by oral presentations (on the last day of the ESSIM School) and written reports (deadline two months after ESSIM).
 
The expected outcome is to provide awareness of the students to take up mathematical modelling as a tool for innovation in improving RD procedures in "Industry".

Preparatory School Course programs

The courses offered during the Preparatory School will be the following:

MCMC Methods and Reliability of Model Predictions

Teacher: Prof. Heikki Haario, Technical University of Lappeenranta
 
Abstract:
The accuracy of modelling is always limited. This is  due to the idealizations in the model itself, and due to errors in measurement that are needed to calibrate the model against real data. The estimation of the impact of noisy data is most often hampered by the fact that the phenomena studied
are nonlinear, while the standard statistical theory employed only is valid for linear models. In this course we will present  recent  computational tools that enable us  properly analyse the reliability of predictions by nonlinear models. 

Program:
  • The classical error analysis of linear and nonlinear models, an overview.
  • Computational statistics: sampling, MCMC (Markov chain Monte Carlo) methods. Efficient practical algorithms.
  • Examples, hands-on exercises with Matlab.
  • A relevant industrial application demonstration.

Downloadable material:

slides

exercises

Data-mining and statistical visualization

Teachers: Prof. Magnus Fontes, Dr. Charlotte Sonesson, Lund University

Abstract:
I will present a collection of statistical and mathematical tools that are useful for the exploration of multivariate data.
The material will be presented in a form that is meant to be particularly accessible to a classically trained mathematician.
We will study spectral methods like principal component analysis, multidimensional scaling, nonlinear Kernel methods, graph based methods and some accompanying parts of statistical hypothesis testing.
Using the presented mathematical framework we will explore some real world high dimensional datasets using statistical and knowledge supported visualizations. In particular we will analyze several different genomewide DNA-microarray datasets.
We will use software from my company Qlucore (see www.qlucore.com) that we will have free access to, but we will also make some experiments writing simple code in R, Matlab or Python.
 
Downloadable material:
 
 

Introduction to Local and Global Optimization for Non Linear Programming

Teacher: Prof. Marco Trubian, Università degli Studi di Milano
 
Program:
  1. Unconstrained Optimization. Optimality conditions. Local and global convergence. Quadratic Programming. A short account of Gradient, Newton, Quasi Newton (BFGS), Coniugate gradient and Trust-Region methods. The Least Square Problem.
  2. Constrained Optimization. Optimality conditions. Penalty functions and barrier methods. Note on the augmented lagrangean method and on the Sequential Quadratic Programming (SQP) technique.
  3. Elements of global optimization. The Ampl modelling language. Syntax and examples. Script files: implementation of global optimization techniques.
References:
J. Nocedal, S.J. Wright. Numerical Optimization. Second Edition. Springer, 2006.
 
Downloadable material:

Level set methods for image reconstruction

Teachers:
Giovanni Naldi, Dipartimento di Matematica, Universita' degli Studi di Milano 
Oliver Dorn, University of Manchester and Universidad Carlos III de Madrid
 
Course Overview:
Image reconstruction and inverse problems play a fundamental role in many industrial applications nowadays. Examples are non-destructive testing, quality control, process and production planning, exploration, amongst others. The same techniques play a fundamental role in medical imaging applications, in Geophysics, and in a large variety of Engineering branches.
In many of these applications, the images which need to be formed from experimental data are complicated, which means they show a certain structure, regions and interfaces. Even worse, often the experimental setup provides only a limited set of data for solving this task of image reconstruction, these data always being noisy to some degree. Mathematically, these issues give rise to very challenging and very interesting problems which need to be solved when tackling these imaging applications.
In the course several recently developed concepts are presented which use a level set formulation of shapes and interfaces for reconstructing complicated images from indirectly obtained data. Several practically important case studies from industrial, medical and geophysical imaging are presented for demonstrating the discussed concepts on real world problems.

The overall structure of the course is planned to be as follows (each lecture of about 90 minutes)
  1. Inverse problems and image reconstruction: A level set approach
  2. Shape evolution and shape optimization techniques
  3. Regularization techniques for reconstructing images with interfaces
  4. Miscellaneous topics for image reconstruction with level sets.
  5. Application of the above techniques to a relevant industrial problem.

Downloadable material:

Classes assignation

Partecipation and final examination to assigned classes is mandatory. Students may also attend the other courses, if they are interested in.

Modelling Week Projects Description

 

Schedule

The Preliminary Summer School and the Modelling Week will be held at the Department of Mathematics, via Saldini 50, Milano.

 
Detailed scheduling of the opening of the Modelling week:
Sunday Sept 5
15.00-16.00 Registration  - Department of Mathematics, via Saldini 50 Milano
16.00-16.20 Welcome and Opening (A.Micheletti) - room: aula Chisini
16.20-17.20 Presentation of the problems (max 5 minutes each) - room: aula Chisini
17.20-17.40 coffee break - room: aula C
17.40-18.30 brainstorming discussion inside the individual groups

Registration

The deadline for registration of students is April 30, 2010.
 
NON ECMI Students: please contact directly the organizers at the address
ecmi.milan@gmail.com
ECMI Students and instructors: please register with your local contact persons, who will contact the organizers:

Autonomous University of Barcelona Frederic Utzet 
Chalmers University of Technology, Göteborg Håkan Andreasson
TU Dresden Antje Noack
TU Eindhoven Martijn Anthonissen
University Joseph Fourier, Grenoble Christophe Prud'homme
University of Jyvaskyla Timo Tiihonen
TU Kaiserslautern Thomas Goetz
Lappeenranta University of Technology Matti Heiliö
Johannes Kepler University Linz Ewald Lindner
Lund University Anders Heyden
Technical University of Denmark, Lyngby Ove Skovgaard
University of Milan Alessandra Micheletti
University of Novi Sad Natasa Krejic
University of Oxford Hilary Ockendon
University of Strathclyde Chris Coles
Tampere University of Technology Robert Piche
University of Tartu Peep Miidla
Norwegian University of Science and Technology, Trondheim Anne Kværnø
Wroclaw University of Technology Agnieszka Jurlewicz
National Institute of Applied Sciences, Rouen Jean Guy Caputo
Carlos III University of Madrid Jose Maria Gambi

Travel infos

Local Organizing Committee

Prof. Alessandra Micheletti (contact person, email: alessandra.micheletti@unimi.it)
Dr. Giacomo Aletti
Dr. Paola Causin
Prof. Marco Trubian

Contact address: ecmi.milan@gmail.com