SEMESTER 5

Semester 5 (UE) nb of hours CM TD TP

Scientific knowledge 65 65
lMathematical methods for
4 60 30 30
engineers
Numerical analysis I 4 70 35 25 10

51
Methods and techniques 106 73

Algorithmie+Unix+Outil du Calcul
Scientifique (Fortran90, 4 70 35 14 21
verification/validation de code)+git

optimisation continue 4 60 25 25 10
Probabilité+ Chaine de Markov+
6 100 46 34 20
simulaton aleatoire

MATHEMATICAL METHODS FOR ENGINEERS

(30HCM, 30HTD, 4 ECTS)

Educational purpose:
· Functions of a complex variable, Fourier transforms, separation of
variables Laplace transforms (10 HCM, 10HTD)
· Vectorial calculus (Green and Ostrogradsky’s theorems) (5HCM,
5HTD)
· Theory of integration (Lebesgue measures, LP spaces, distributions
and basic properties , Sobolev spaces ) (15HCM -15HTD)

NUMERICAL ANALYSIS I
(35HCM, 25HTD, 10HTP, 4 ECTS)

Educational purpose : To acquire the basic numerical methods used to

solve linear and non-linear problems ( with finite difference
dicretizations)
Content :
1. Finite difference discretization : Basic principles (Taylor’s formula), 2D
and 1D examples (explanation of the numbering 2D->1D, use of
maple), Richardson’s extrapolation.
2. Direct methods: LDU factorization (Existence and uniqueness
conditions of the decomposition, matrix conditioning, impact of
conditioning on the error, pivoting technique, computation of the
determinant),profile conservation ( application to periodic boundary
conditions), Schur complement and its properties.
3. Numerical interpolation and integration: quadrature formulas (Newton-
Cotes, Romberg’s methods), Gibbs phenomenon, orthogonal
polynomials , Hermite’s interpolation, (Divided differences formula and
associated error estimation) Interpolation using rational functions
(Padé approximant)
4. Newton’s methods/fixed point iteration: Basic priciples (fixed-point
method , Newton’s method),Affine invariance theorem, quasi -Newton
methods
5. Basic iterative methods: splitting methods: (Jacobi, Gauss-Seidel,
,SOR), General convergence theorems (based on spectral radius,
optima SOR), Krylov’s methods: (principles, GC, GMRES),
Convergence theorems based on eigenvalue distribution
6. Solution methods for ODE’s: One-step methods/Theory (consistency,
stability, convergence impact of round-off errors, Grownwall’s lemma,
Euler methods, implicit, explicit, mid-point, adaptive step size
methods, stiff problems), Introduction to multistep schemes (BDF)
References:
J.Stoer & R. Bulirsh: Introduction to numerical analysis, Springer text in
applied Math. 15
G.H. Golub & C.F. Van Loan: Matrix computation, J. Hopkins University
press
E. Hairer & G. Wanner : Solving Differential Equations I, Springer series
in Comput. Math 14
Y. Saad : Iterative methods for sparse linear systems, SIAM

COMPUTATIONAL TOOLS
(35HCM, 14HTD, 21HTP, 4 ECTS)

Educational purpose :
Mastery of basic of algorithmics,Unix command of version control
systems, imperative programming, code verification and validation
techniques
Prerequisite : Numerical analysis I
Content:
1. Algorithmic structures:identifiers arithmetic and boolean expressions,
statements: syntax and semantics,standard algorithmic constructions.
(6HCM+6HTD)
2. Unix: on line help, file system environment variable,files handling
commands makefile and shellscripts creation, . (3HCM/3HTP)
3. Version control systems:Introduction to GIT: create a repository
,commit sources,retrieve sources from a repository,create
development branches, go back to previous version. (2HCM/2HTP)
4. Finite arithmetics of computers and their consequences : computation
conditioning, backward or a posteriori analysis. (2HCM/2HTD)
5. Imperative programming: Fortran 90, Tables,complex data structures,
procedures and functions, recursion, dynamic allocation, module.
Implementation of numerical analysis methods I(CSR stocking,
iterative methods and Krylov’s method). Introduction to BLAS and
LAPACK scientific libraries . (16HCM/16HTP)
6. Code verification: software quality certification (static and dynamic
analysis, unit testing, consistency and convergence, formal and
observed accuracy order , manufactured solutions methods
.(4HCM/4HTD)
7. Code validation: good practices guide, entrained cavity
example(2HCM/2HTD)
Support software : Fortran90, BLAS, LAPACK, Maple
References:
R. Séroul: Programming for Mathematicians, springer 1995
M. Metcalf, J.Reid, M.Cohen: Fortran 95/2003 explained, Oxford
University press, 2007
P. Roache: Verification and Validation in Computational Science and
Engineering, Hermosa publishers, 1998

CONTINUOUS OPTIMIZATION
(25h CM, 25h TD, 10h TP ; 4 ECTS)
Educational purpose :
To recognize an optimization problem : to deal with the existence of
solutions to the problem and provide (an) approximate value(s) of the
solution(s) using appropriate numerical methods
Content :
Optimization without constraints and with equality or inequality
constraints . Extremum conditions.Solution of F(X)=0 using Newton and
quasi-Newton methods; application to optimization. Lagrange multipliers,
saddle point, duality.
Software support : Matlab.

PROBABILITY AND SIMULATION OF RANDOM EVENTS

(50h CM,24h TD,16h TP, 6 ECTS)
DESCRIPTIVE STATISTICS WITH R - 4h CM , 6h TP
Educational purpose
By the end of the course, students are expected to be able to analyze a
random variable using standard parameters as well as the interaction
between 2 such variables. They will also master the R software.
Content
Satndard parameters : mode, mean, median, quantiles, range, standard
deviation, kurtosis, skewness
Relevant graphs : boxplots, histograms, pie charts ;
Relationship between 2 variables : covariance, correlation,
cindependence ;
Software R : handling of data, obtain basic statistical parameters and
graphs .
Support software : R
References
Cornillon, P.A.et al., Statistiques avec R, Presse Universitaire de France,
2010.
Grais, B., Méthodes statistiques, Dunod, 2003.

PROBABILITY THEORY - 30h CM, 30h TD

Educational Purpose
By the end of the course, students are expected to be able to handle and
find out probability distributions and to identify the common ones .
Besides they are also expected to be conversant enough with stochastic
convergence and conditioning and related theorems to be able to apply
them in the various subsequent probability and statistics courses.This
course is essential to the understanding of courses dealing with random
modeling.
Prerequisite: Mathematical methods for engineers.
Content
Conditional probability, independence ;
Random variable : common distributions , variable transformation,
quantiles, moments ;
 Random vectors, gaussian vector ;
Laplace transform, characteristic function ; conditional expectation ;
Various types of convergence, law of large numbers, central limit
theorem ..
References
Foata, D. & Fuchs, A. Calcul des probabilités, cours, exercices et
problèmes corrigés, Dunod, 1998.
Jacod, J. & Protter, P., L’essentiel en probabilités ou bien les bases de la
théorie des probabilités., Cassini, 2003.
Billingsley, P., Probability and measure,Wiley, 1995.
Grimmet, G.R.& Stirzaker, D.R., Probability and Random Processes,
Oxford Science Publications, 1992.

RANDOM SIMULATION PART I - 10h TP

Semester 6 nb of hours CM TD TP
Scientific knowledge 60 60
Statistique inférentielle 4 60 30 30

Modèles Mathématiques pour système multi-

physiques 4 60 30 30
(Electro/thermo/automatique/mecaflux/Elasticité)

Methods and techniques 104 20 116

Genie logiciel (Programmation Orienté
4 60 28 4 28
Objet+UML Lite)
Recherche Opérationelle et CAO 3 60 25 10 25

Analyse de données (SAS) et base de donnée 4 72 34 38

Simulation aléatoire en statistique (logiciel R) 3 48 22 26

Educational purpose
By the end of the course, students should be able to simulate random
variable distributions and will have put their knowledge of probability
theory into practice.
Prerequisite: Course in probability.
Content
Random variables simulation :inverse distribution function,
acceptance/rejection...
Monte Carlo methods ;
Hands-on exercises relating to the course in probability
Support software : R
References
Bouleau,N., Probabilités de l’Ingénieur, variables aléatoires et simulation,
Hermann, 2002. Devroye, L., Non-Uniform Random Variate Generation,
Springer, 1986.
Robert, C. et Casella, G., Monte Carlo StatisticalMethods, Springer-
Verlag, 2004.

MARKOV CHAINS - 12h CM, 4h TD, 4h TP

Educational purpose
This course is meant to teach the fundamentals of markovian
phenomena modeling.By the end of the course, students will have been
given the opportunity to put the above concepts into practice (
reliability,waiting lines, genetics, economics..)
Prerequisite Probability theory, Random simulation Part I.
Content
Definition, transition matrix and standard examples.
Chapman-Kolmogorov equation and conditioning.
Classification of states, periodicity, waiting time, recurrence and
transience.
Steady-state distribution and limit theorems.
Support software : R
References
Pardoux, E., Processus de Markov et applications : Algorithmes,
réseaux, génome et finance. Dunod, 2007.
Foata, D. & Fuchs, A., Processus stochastiques - Processus de Poisson,
chaînes deMarkov et martingales, Cours, exercices et problèmes
corrigés, Dunod, 2002.
Bercu, B & Chafai, D., Modélisation stochastique et simulation, Dunod,
2007.
 Semester 6
INFERENTIAL STATISTICS
( 30h CM, 30h TD, 4 Ects)

Educational purpose:By the end of the course, students should master

statistical modeling besides being be able to build estimators and study
their properties.They should have become proficient in the construction
of confidence intervals and in the use of standard statistical tests,
whether parametric or not, both from a theoretical and practical
perspective.
Prerequisite: Probability theory
Content
Statistical models ;
Parametric estimation : insertion method, qualities of an estimator ;
¦ likelihood estimation : maximum likelihood, Fisher information,
properties of the maximum likelihood estimator ¦ confidence interval ;
¦ statistical tests :associated risks, optimality, asymptotic test,likelihood-
ratio test, independence and suitability test ;
¦ Gaussian model : Cochran’s theorem, multiple linear model, Fisher’s
test, means comparison test ;
¦ Applications : statistical methods in Markov chains
References
¦ Rivoirard, V. & Stoltz, G., Statistique en action, Vuibert, 2009.
¦ Fourdrinier, D., Statistique inférentielle-Cours et exercices corrigés,
Dunod, 2002.

MULTIPHYSICAL MODELING
(ELECTRO/THERMO/AUTOMATICS/FLUIDMECHNICS/ELASTICITY
(30 h CM, 30h TD, 4 ECTS)
Educational purpose:
1. Understanding the mathematical models of multiphysics systems, the
classification of variables and the relationships between them
2. Hamiltonian formulation of dynamic models of open physical systems
and of their properties.
3. Covariant formulation of conservation laws and spatial discretization
methods, while keeping their symplectic or Dirac structure .
4. Controllability and stabilization using Lyapounov’s methods
5. Simulation and control using softwares (Matlab®, Scilab® ...).
Prerequisite: Basic knowledge of differential and vector calculus ,
Fourier and Laplace transforms.
Content
The course consists of 4 parts.
· The first part deals with macroscopic multi-physics systems which
are governed by systems of conservation laws with source terms
(balance laws) .A structural classification of such systems is given
according to the nature of their variables (storage variables , fluxes,
driving forces, thermodynamic state variables) and of the
relationships between them.This structure will be highlighted by
different applications pertaining to heat, mass and momentum
transfer, to elastodynamics, fluid mechanics and electromagnetism.
·
· The second part addresses the Lagrangian and Hamiltonian
formulations of open multi-physics systems and of their
coupling.Finite and infinite dimensional port-Hamiltonian systems
with dissipation will be introduced as well as their interconnection
and the associated composition of Dirac structures.Properties of
such systems are discussed: dynamic invariants, dissipativity,...
·
· The third part discusses the properties of the systems of PDE’s
associated with the above models and introduces both the
differential forms and the covariant formulation of the conservation
laws. It also deals with ways to adapt finite element discretization
schemes and pseudo-spectral methods preserving the hamiltonian
structure of dynamical systems.
· The fourth part which is devoted to Continuous automatic control,
presents controlled systems and their structural properties, along
with feedback command and state variable feedback control.The
state approach issue is essentially dealt with, relying on a system
of 1rst order ODE’s, on the controllability and observability
properties and on the state variable control synthesis.Where non-
linear systems are concerned, control Lyapunov functions are
introduced since they are particularly suited to dissipative port-
Hamiltonian systems.
 · References:

RANDOM SIMULATION PART II

(22H CM, 26H TP, 3 ECTS )
Educational purpose:
This course first aims at putting the main results of Inferential Statistics
into practice but , in a second step, also touches upon advanced
statistical methods ( non-parametric statistics for example), which prove
to be very useful, such as Bootstrap’s methods.
Prerequisite: Probability theory, Inferential statistics, Random simulation
Part I
Content
· Simulations related to the course in Inferential Statistics : point and
interval estimation, parametric and non-parametric tests (CM 2h, TP
10h) ;
· parametric and non-parametric Bootstrap (ex :Estimation algorithms
for Bias/Variance/Mean squared error/Confidence intervals)(CM 10h,
TP 10h);
· Non parametric statistics (density estimation, regression), rank test
(CM 10h, TP 6h).

Software support : R
References
Silvermann, B.W., Density Estimation for Statistics and Data Analysis,
Chapman & Hall, 1998.
Tsybakov, A. B., Introduction à l’estimation non-paramétrique, Springer,
2004.
Efron B.& Tibshirani, R. J., An introduction to the bootstrap, 1994.
Van der Vaart, A.W. Asymptotic Statistics Cambridge university press,
1998.

DATA ANALYSIS AND DATABASE

(34h CM, 38h TP, 4 ECTS)

DATA ANALYSIS AND CLASSIFICATION - 24h cours, 28h TP en

SAS
 Educational Purpose
By the end of the course, students should be able to put the standard
data analysis methods into practice and should have gained some
proficiency in the use of the SAS software.
Prerequiqite : Descriptive statistics using R.
Content
· Introduction to SAS (Data stage, procedure stage, graphical
procedures and first elementary stastics procedures ) (CM 10h, TP
10h);
· Data analysis (CM 16h, TP 18h)
o Principal component analysis(PCA) ;
o Factorial correspondence analysis (FCA)
o Multiple correspondence analysis (CAM);Classification method
;Discriminant analysis (decision and exploratory analysis)
o Support software : SAS
References
Droesbeke, J.J & Fichet, B. & Tassi, P., Modèles pour l’analyse des
données multidimensionnelles, Economica, 1992.

Semester 7 nb of hours CM TD TP
STAGE S7
30 STAGE 1 Septembre 15 Janvier

Escofier, B. & Pages, J., Analyses factorielles simples et multiples,

Dunod, 1998.
Husson, F. & Le, S. & Pages J., Analyse de données avec R. Presses
Universitaire de Rennes, 2009.

DATABASES - (10h CM - 10h TD)

Educational purpose :
To acquire the basics of database manipulation in order to apply them to
data-mining and risk analysis.
 Prerequisite: Basic knowledge of algorithmics and statistics.
Content :

Semester 8 CM TD TP HP
Scientific knowledge 26 14 20
Modelisation en statistique 4 60 26 14 20

Methods and techniques 125 85 60 30

Méthode de discretisations 5 85 38 32 5 10
Analyse Numerique II 4 70 25 20 15 10
Problème instationnaires 5 75 30 15 15
Serie temporelle et modèle de durée 4 70 30 20 10 10

Storage of huge amounts of complex data/introduction to its principles–

Data search and sorting within a large data set.Database organization.
Object databases. Basic knowledge of SQL.
Support software : Access, Spad.

DISCRETE OPTIMIZATION & CAD

(25h CM, 10h TD, 25h TP ; 3 ECTS)

DISCRETE OPTIMIZATION (10h CM, 10h TD, 10h TP )

Educational purpose :
To enable students to implement algorithmic solutions to graph type
problems.
Prerequisite : Basic algorithmics
Education purpose :Linear programming, simplex. Integer
programming. Assessment and separation procedures,heuristic
procedures. Annealing and genetic algorithms . Graphs ; pathfinding,
flows, cover tree, maximum coupling. Introduction to scheduling
methods
Support software : Matlab

CAD (15h CM, 15h TP )

Educational purpose:
To acquire the basic concepts involved in geometry manipulation in CAD.
To learn about the different existing design models.To implement a CAD
code used in industry. An upstream intervention stresses the challenges
of CAD and its integration within companies.
Prerequisite: Basic knowledge of curves and surfaces. Basic numerical
methods and Matlab numerical applications.
Content:
Upstream : Introduction and role of CAD in the iterative design process;
Presentation of current CAD systems and of the main suppliers of CAD
and PDM.Different design models depending on market shares; Outline
of current issues.
Geometric modeling of curves and surfaces : Bézier model : definition of
curves ; De Casteljau’s algorithm and other algorithms (subdivision,
degree elevation…) ; Bézier patches ; Sticking patches together and
geometric continuity. Shortcomings of the model. B-Spline Model, De
BOOR’s evaluation algorithm, geometric properties and limitations.
NURBS Model : curves and surfaces. Aspects of « algorithmic
geometry» : shape triangulation, search for convex envelopes for the
calculation of intersections. Other geometric design models (CSG,
BREP). Exchange formats.
CAD practice.
Support software : CATIA V6 ou ProEngineer.

SOFTWARE ENGINEERING
(28h CM, 4h TD, 28h TP ; 4 ECTS)
Educational purpose: To have a methodological tool available to
implement the object approach (UML2.1.1). To be able to implement the
concepts of object-oriented programming using the C++ language : to be
able to build a simple application and to contribute to a large and
complex one.Introduction of the object concept in scientific computation.
Prerequisite: Tools of scientific computing , Numerical analysisI.
Content:
°Management of a complex application code : object solution using UML
2.1.1 (4HCM,2HTD,2HTP) UML methodology concepts ( OMG norm
(Object Management Group)). Static views of a system:object, class,
component, deployment diagrams. Dynamic views of a system:
collaboration, sequence,state-transition, activity diagrams
o Object-oriented programming for scientific computing: (16HCM,
16HTP). Basics of POO: object, class, polymorphism, dynamic link,
C++ programming : class, operators, friendship concept, generic
programming : (template).Simple and multiple Inheritance . STL and
scientific libraries implementation (4HCM/6HTP).Structure of
advanced data for scientific computing : valarray, map, vector.
Introduction to the Template Numerical Toolkit library (lapack++,
IML++, SparseLib++, mv++.). Inclusion of new digital features into a
complex object-oriented code designed for scientific computing
(hydrology, process engineering,…).
°Development of human-machine interfaces (2HCM,2HTD, 2HTP). Use
of the Qt library and graphism .
o Code optimization, debugging tolls (2HCM, 2hTP)
Support softwares : Environnement Eclipse Environment,
OMONDO/BOUML, STL library, Boost library, GSL 1.9, Template
Numerical Toolkit, Java Sun et BEA Logic compilers.

Semester 7
Semester 8

STATISTICAL MODELING
(28 HCM, 14H TD, 18H TP, 4 ECTS)
BAYESIAN MODELS - 8h CM, 8h de TD/TP
Educational purpose
Students are expected to be able to implement a bayesian methodology
for standard modelings : from the choice of a prior probability distribution
to the construction of a confidence interval.Practical applications will be
run using the R software.
Prerequisite: Markov chains, Inferential statistics, Random simulation
Parts I et II.
Content
Markov chain Monte Carlo methods (MCMC);
Prior and posterior probability distributions (explicit or through MCMC) ;
Bayesian inference : bayesian risk, estimation,confidence tests and
intervals.
Support software : R
References
Robert, C. P., L’analyse bayésienne, Sringer, 1992.
Robert, Ch., Méthodes deMonte Carlo par chaînes deMarkov,
Economica, 1996.Gilks,W. R. & S. Richardson, & D. J. Spiegelhalter,
Markov chainMonte Carlo in practice, Chapman and Hall, 1996.

REGRESSION MODELS - 20h CM, 24h TD/TP

Educational purpose
By the end of the course, students should be able to perform a linear
regression analysis,to critically analyze a software output, to treat
atypical points , to choose a suitable model and to master the variance
and covariance analysis models. They will be able to implement the
various models using SAS and R.
Prerequisite Random modeling of semesters S5 et S6.
Content
 Simple and multiple linear regression : model,least square estimators,,
gaussian model, confidence interval, tests, prediction,residual analysis,
model selection ;
Study of special cases : ANOVA, ANCOVA ;
Support softwares : R et SAS
References
Azaïs J-M. & Bardet J-M., Le modèle linéaire par l’exemple Régression,

Semester 9 nb of hours CM TD TP HP
Scientific knowledge 45 45
Modelisation Mathématique-Galerkin
5 90 45 45
discontinu

Methods and techniques 108 51 51 20

Calcul Haute performance 4 70 35 35
statististique des processus 3 50 26 20 4
Modélisation statistique avancée 4 60 32 16 12
Projet 6 80 15 15 50
option Big Data 15 15
option contrôle 15 15

analyse de la variance et plans d’expérience illustrés avec R, SAS et

Splus, Dunod, 2006.
Matzner-Løber, E. & Cornillon, P-A., Régression, Théorie et applications,
Springer, 2007.Weisberg, S., Applied Linear Regression,Wiley, 2005.
Droesbeke, J.-J., Fine, J., Saporta, G., Plans d’expériences -
Applications à l’entreprise, Technip, 1997.

DISCRETIZATION METHODS: (38h CM, 32h TD, 5h TP ; 10h HP, 5

ECTS )

Educational purpose:
To learn how to numerically approximate PDE problems (elliptic, linear
parabolic and hyperbolic) using the finite element method (FEM)
Application to structural and fluid mechanics: linear elasticity, Navier-
Stokes,..
Prerequisite: Numerical analysis I,Mathematical methods for engineers

Content:
Examples of elliptic,parabolic and hyperbolic problems. Variational
formulation of elliptic boundary value problems. Description of Sobolev
spaces. Lax-Milgram theorem.
Finite element approximation using the examples of diffusion, elasticity,
transmission, diffusion-convection problems… Finite element spaces:
examples of 1D, triangular, rectangular elements, and extension to 3D.

Stability and convergence analysis for elliptic problems.

(Smoothness of the exact solution and its finite element approximation).

Application to a few elliptic problems in continuum mechanics

(Elasticity, Stokes).
(Smoothness of the exact solution and its finite element approximation)

Fundamentals of FEM implementation.Practice of a FEM code actually

used in R&D.

Mixed formulation of elliptic problems and Raviart-Thomas type finite

element approximation.

Mixed formulation of incompressible fluid flow problems. Stokes

problem.Stability and convergence analysis.. Inf-Sup condition (LBB).
Usual finite element spaces and associated convergence results.
Pressure treatment.

Solution of saddle-point problems: uzawa et penalty methods.

Mixed formulations in solid mechanics for plate problems (Hellinger-
Reissner-Mindlin, Hu-Washisu approaches,...).
Consideration of the role of non-linearity (convection), transport
processes, reaction : Stabilization using Upwind, SUPG, GALS, mass-
lumping methods…
Support software :
COMSOL Multiphysics / CAST3M (CEA)
References:
P.G. Ciarlet: The finite element method for elliptic problems. North-
Holland 1978.
A. Quarteroni et A. Valli : Numerical Approximation of partial differential
equations, Springer-Verlag, Berlin, 1994.
O. Pironneau: Méthode des éléments finis pour les fluides. Collection
Mathématiques appliquées Masson, 1988. O.C. Zienkiewicz: The finite
element method in engineering sciences, Mac Graw Hill, 1971.
B. Lucquin et O. Pironneau: Introduction au calcul scientifique. Collection
Mathématiques appliquées Masson, 1995.
NUMERICAL ANALYSIS II
(25h CM, 20hTD, 15h TP, 10h HP, 4 ECTS )
Educational Purpose: To master the methods for solving the large
linear systems resulting from the discretization of PDE’s and for
calculating the eigenvalues of large dimensional matrices.Basic
knowledge of domain decompostion methods.
Prerequisite : Numerical analysis I and Computational tools
Content :
· Methods for eigenvalue computation: power method, reduction of a
matrix to a simpler form (Hessenberg, Givens rotation, Householder
methods,reduction of a hermitian matrix to tridiagonal form), QR and
Davidson-Jacobi methods, Singular value decomposition (symmetric
case)
· Krylov accelerated methods:
· ILU preconditioning, geometric and algebraic multigrid
preconditioning, deflation methods, domain decomposition methods:
· Primal and dual (Feti) Schur complement methods , Schwarz type
domain decompositon methods. Convergence acceleration (ORAS,
Aiken-Schwarz), DDM as a preconditioner of Krylov methods (BPS,
RAS,...)
· Solving methods for stiff ODE/ADE problems: Shooting methods,
Runge Kutta methods, Symmetric integration schemes , symplectic
schemes for hamiltonian systems, ADE index , problem-solving
techniques.
References:
J.Stoer & R. Bulirsh: Introduction to numerical analysis, Springer text in
applied Math. 15
G.H. Golub & C.F. Van Loan: Matrix computation, J. Hopkins University
press
E. Hairer & G. Wanner : Solving Differential Equations II, Springer series
in Comput. Math 14
Y. Saad : Iterative methods for sparse linear systems, SIAM

UNSTEADY PROBLEMS
(35hCM, 25hTD, 15hTP , 5 ECTS)
Educational purpose :To become conversant with computing
techniques, consistency and stability (dispersion/diffusion) analysis and
the conservation of schemes derived from a differential operator-based
theoretical approach. The discretization schemes used here are
essentially restricted to finite difference and volume methods.
Prerequisite : Mathematical methods for engineers I,Basic numerical
methods
Content :
a) Numerical schemes for evolution PDE’s of first and second order with
respect to time. Heat, advection-diffusion, Burgers’, wave and reaction-
diffusion equations. Finite difference schemes : stability analysis (von
Neumann’s and Fourier’s methods, energy method), dispersion and
diffusion analysis. Error and consistency order of finite difference
schemes. CFL condition. Upwind schemes.
b) Conservation laws, applications, difficulties, examples : advection,
Burgers’ and road traffic equations. Systems of conservative laws : wave,
Euler, d Navier-Stokes, and Saint-Venant equations. Standard solutions
and method of characteristics: linear and non-linear case. Limitations of
the method of characteristics which make it necessary to introduce the
more general concepts of weak and entropic solutions. Study of a
conservation law : differential and integral forms , weak solutions,
Rankine-Hugoniot relationships, entropy concept, entropy jump,
expansion waves.Riemann problem.
Solving Riemann problem for non linear laws of conservation.
Conservative scheme concept. Conservative methods for non-linear
problems : conservative methods, consistency,discrete conservation ,
Lax-Wendroff theorem,entropy condition. Godunov’s scheme. Non linear
stability :TVD and monotonic methods.
c)Upwind finite volume methods.
Approximate Riemann solvers : general theory. HLL and HLLC,
Rusanov, Roe, Osher, WAF Riemann solvers. Introduction of such
methods using a recent more comprehensive and cost-effective PVM
approach,.
High order methods and TVD schemes: reconstruction methods,
MUSCL approach, generalized Riemann problem, monotonic schemes
 and precision, flux and slope limiters. Extension of TVD methods. Source
term associated problem. Well balanced schemes.
Support software : Matlab
References :

Semester 10 nb of hours CM TD TP
INTERNSHIP S10
30 End of program internship 15 Février 30 Septembre

* Grégoire Allaire, Analyse numérique et optimisation, Les Editions de

l'Ecole Polytechnique, 2006
* Randall J. LeVeque, Numerical methods for conservation laws,
Birkhäuser, 1992
* E. Godlewski, P.A. Raviart, Numerical approximation of hyperbolic
systems of conservation laws, Applied Mathematical Sciences, 118,
Springer, 1996
* TIME SERIES ET DURATION MODELING
( 30hCM, 20hTD, 10hTPn 10hHP, 4Ects)

TIME SERIES - 20h CM, 14h TD 6h TP

Educational purpose
The course aims at presenting statistical treatment methods related to
time series : smoothing,deseasonalization and forecast.By the end of the
course students should have developed an awareness of the potential
application of time series in industry, in the economy, .. and be able to
analyze them using software outputs.
Prerequisite: Random modeling courses of semesters 5 and 6,
Regression models.
Content
Descriptive analysis of time series (seasonal decomposition, exponential
smoothing) ;
Random modeling of a time series : second order process, stationarity,
autocovariance function, autocorrelation function, partial autocorrelation
function, spectral density ;
Univaried processes :MA, AR, ARMA, ARIMA, SARIMA ;
Practice of SARIMA models( Box-Jenkin methodology) : identification,
estimation, verification, validation, comparison.
Support software : R et SAS
References
¦ Brockwell, P. & Davis R., Introduction to Time Series and Forecasting,
Springer, 1996.
¦ Bosq D., Lecoutre J-P., Analyse et prévision des séries chronologiques.
Méthodes paramétriques et non paramétriques,Masson, 1992.
¦ Aragon, Y., Séries temporelles avec R : Méthodes et cas, Springer,
2011.

DURATION MODELS - 10h CM, 6hTD, 4hTP

Educational purpose
By the end of the course, students should be familiar with the vocabulary
and the tools used in duration modeling as well as with its main
applications to reliablity , health care issues,.. They are expected to be
familiar with the models and the associated underlying assumptions as
well as with the commonly encountered problems. They should also be
able to calculate standard estimators and interpret standard tests, using
the R and SAS softwares
Prerequisite: Random modeling courses of semesters 5 and 6
Content
Duration models ;
Parametric and non-parametric modeling
Regression models ;
Bayesian approach.
Support software : R or SAS
References
Droesbeke, J.J. & Fichet, B. & Tassi, P., Analyse Statistique des durées
de vie. Modélisation des données censurées
Klein, J.P.&Moeschberger,M.L., Survival Analysis : Techniques for
Censored and TruncatedData, Springer, 2003.

Semester 9
MATHEMATICAL MODELING-DISCONTINUOUS GALERKIN
(45HCM ,45 HTP, 5 ECTS)
DISCONTINUOUS GALERKIN METHODS (CM 15, TP 15)
Educational purpose :
By the end of the course, students should be able to discretize PDE
systems using the discontinuous Galerkin method and should be
conversant with its properties, in particular from a numerical point of
view.
Prerequisite : Mathematical methods for engineers , Finite elements I
Content:Definition of the method’s specific functional framework .
Discontinuous variational formulations, broken Sobolev spaces,non-
compliant error analysis. Reminders on orthogonal polynomials(for
approximation spaces). DG methods for scalar and non-linear
conservation laws : weak formulation, well-posed problem, stability,
convergence, error computation, inf sup condition. RKD methods. How to
introduce diffusion operators in the DG methods(symmetric interior
penalty methods, ...). DG methods will be shown to be mixed methods
that are stabilized for diffusion problems. Numerical fluxes
approximation:centered and upwind schemes, examples and
applications.Use of non conforming meshes. Unsteady
problems.Application to Stokes, Navier-Stokes and Saint-Venant
equations.
References :
D.A. Di Pietro, A. Ern, Mathematical aspects of discontinuous Galerkin
methods, Springer, 2012.
B. Rivière, Discontinuous Galerkin methods for solving elliptic and
parabolic equations: Theory and implementation. SIAM, 2008

MODELING (CM 30, TP 30)

Educational purpose :
To acquire a command of all the problems involved from the modeling of
a physical problem to its numerical solution and of the underlying
mathematical tools.Students are also expected to be able to combine
and implement the discretization methods they have been taught during
the first two years of their curriculum.
Prerequisite : Courses in Numerical methods
Content:
In this course, the modeling process is tackled all the way from its
physical formulation to its numerical solution. The program consists of
several independent modules:
- Modeling in biology : reaction-diffusion equations describing animal
coat pattern formation. Analysis of solutions.
Solving non linear models using Newton and Gear’s methods.
- Modeling of the valuation of financial products ( options, futures).The
Black and Scholes equations are derived from brownian motion and
discretized using finite difference and finite element methods.
- Modeling in fluid mechanics : the main equations (momentum and
continuity) are derived from conservation laws. Lagrangian vs. Eulerian
description.The turbulence problem.
Application to various problems in geophysics.
- Modeling of dynamical systems, bifurcations, non linear analysis.

Support software : FreeFem++

References : They will be provided during the course in the form of
articles in journals, book chapters, etc..
HIGH PERFORMANCE COMPUTING
(35HCM, 35HTP, 4 ECTS)
Educational purpose :
To acquire the reflex of conceptualizing distributed programming, of
writing a new module compliant with the standards of a given code, and
of validating and verifying a code.
Prerequisite : Numerical analysis I and II, Tools of scientific computing
and Software engineering.
Content :
Introduction to message exchange libraries (MPI) ,compilerdirectives
programming (OpenMP) and to computation using GPGPU (opencl).
Modeling of distributed architectures, code performance assessment for
a given architecture.
High performance computing methods suited to the programming
models : pipeline, understructuring, domain decomposition (Schur, Feti,
Schwarz), decomposition of operators, decomposition in function
spaces,distributed DFT, parallel Krylov methods. Code and/or PDE
coupling methods.
Code validation and verification procedure .
Software support :
MPI, openMP, Opencl, Fortran90, Petsc, Scalapack.
ADVANCED REGRESSION MODELS
(24h CM, 14h TD, 12h TP, 4 ECTS)
Educational purpose
This course aims at further developing knowledge acquired in the
Regression models course .Different models will be presented and
applied to concrete situations. Students should eventually be able to
implement and interpret such models: mixed and generalized linear
models .They will also have some basic knowledge of experimental
design and of non parametric modeling.
Prerequisite: Random modeling of semesters S5 et S6, Regression
models.
Content
Mixed models ;
Experimental designs ;
Generalized linear models , logistic regression in particular (link to the
classification) ;
PLS regression, Ridge ;
Non- and semi parametric regression models
Software support : R et SAS
References
Dobson, A.J., An introduction to generalized linear models, Chapman &
Hall, 1990.
Antoniadis, A. & Berruyer, J. & Carmona, R. Régression non linéaire et
applications. Economica, 1992.
Verbeke, G. &Molenberghs, G. LinearMixedModels in practice, Springer,
1997.
Ruppert,D. & Wand, M. P. & Carroll, R. J., Semiparametric Regression.
Cambridge Series in statistical
and Probabilistic, 2003.

PROCESS AND RISK STATISTICS

(32h CM, 20h TD, 8h TP)

PROCESS STATISTICS - (20h CM, 12h TD, 8h TP)

Educational purpose
By the end of the course, students should be conversant with the
modeling of the main time-dependent random phenomena occuring in
industry and biology.... These modelings will be studied both from a
probabilistic and a statistical point of view.
Prerequisite Random modeling of semesters S5 et S6.
Content
Introduction to processes ;
Independent increment processes and brownian motion
Poisson process ;
Birth and death process, renewal process ;
Waiting lines
Software support : R
References
Foata,D. & Fuchs, A. , Processus stochastiques - Processus de
Poisson, chaînes deMarkov et martingales, Cours, exercices et
problèmes corrigés, Dunod, 2002.
Bercu, B & Chafai, D., Modélisation stochastique et simulation, Dunod,
2007.

RISKs - 12h CM, 8h TD

Educational purpose
How to evaluate the probability of sparse events ?How to determine the
height of a dyke for the probability of any overflow to be smaller than
some regulatory value?How to set the rate of a reinsurance premium?
The course is intended to provide the probabilistic and statistical tools
needed to answer the above questions.Attention will also be paid to the
numerical implementation of the methods proposed in R.
Prerequisite : Random modeling of semesters S5 et S6.
Content
Introduction to extreme value theory.Examples.
Maximum and threshold overrun modeling .
Evaluation of small probabilities et extreme quantiles .
Sotware support : R
References
Resnick, S. I., Heavy-tail phenomena. Probabilistic and statistical
modeling, Springer Series in Operations Research and Financial
Engineering. Springer, 2007.
De Haan, L. & Ferreira, A., Extreme value theory. An introduction.
Springer New York, 2006
Beirlant, J. & Goegebeur, Y. & Segers, J. & Teugels, J., Statistics of
Extremes, Theory and applications. Wiley, 2004.

PROJECT
(15h CM, 15h TD, 50h HP)
Educational purpose
Implementation of the knowledge gained over the course of the program
by carrying out of a supervised long term project. Students are offered
optional further training in finance or control.

Semester 10