SYLLABUS
National school on artificial intelligence
Level: Third year
Semester: S5 University year: 2023/2024
Identification of teaching module
Module: Stochastic Modeling and Simulation
Teaching Unit: Methodological
Number of Credits: 5 Coefficient: 3
Total number of hours per week:
o Classes, Tutorials and Labs: 2h, 1h30 and 30 minutes
Responsable of teaching module
Instructor: Dr. Nawel Arrar Remita, Senior lecturer
Email:
[email protected]Meets course: Sunday 10h10-12h10 Amphi 1 - Section1
Monday 11h40-13h40 Amphi 1 - Section2
Teaching assistant: Abdelkader Amraoui [email protected]
Page 1 / 5
Description of the teaching module
Prerequisites: Probability, statistic, inference and transition graphs.
General Objective: This course is aimed at computer systems students and particularly of
artificial intelligence. The field of modeling and stochastic processes has been around for
almost a century, and can be used to save system designers a considerable number of hours of
trial and error, while improving system performance. Stochastic modeling can be used in
conjunction with simulation to help guide simulation, reducing the number of cases that need
to be explored
Learning Objectives: This course is divided into three parts: the first part is dedicated to
stochastic modeling, with a review of Markov chains on discrete and continuous spaces and
their matrix interpretations. The second part is dedicated to stochastic simulation, giving an
overview of techniques for generating and evolving sequences of pseudo-random numbers,
and their use in the production of random variables and vectors obeying any law. Another
stochastic simulation idea, called the Monte Carlo method, is dedicated to solving purely
deterministic problems, such as the evaluation of multiple integrals or combinatorial
optimization. We also present classical Markov Chain Monte Carlo algorithms MCMC.
In the third and last part, we present some areas of application in Bayesian statistics and
hidden Markov chain models.
Course Content
Part 1: Stochastic Processes and their Classification
o Discrete time Markov chains
o Continuous time Markov chains
Part 2: Introduction to Stochastic Simulation
o Pseudo random number generator
o Simulation of random variables
o Simulation of random vectors
o Monte Carlo methods and variance reduction methods
o Simulation of stochastic processes (MCMC)
Part 3: Probabilistic Machine Learning
o Bayesian networks
o Hidden Markov Models
Page 2 / 5
Modalités d’évaluation
Nature du contrôle Pondération en %
Final Exam 60
Tutorial (tests) 12
Labs 10
Mid term 15
Group work
Assiduity 3
Total 100%
Références & Bibliographie
Textbook (Référence principale) :
Titre de l’ouvrage Auteur Éditeur et année d’édition
1. Précis de recherche - R. Faure, B. Lemaire, DUNOD
opérationnelle C. Picouleau
2. Modélisation et simulation - Amar Aissani OPU
3. Recherche opérationnelle pour - H. F. Hêche Presses Polytechniques
Ingénieurs 2
4. An Introduction to Stochastic - M. A. Pinsky et S. 4th ed. Academic Press, 2011
Modeling. Karlin
Les références de soutien si elles existent :
Titre de l’ouvrage (1) Auteur Éditeur et année d’édition
Titre de l’ouvrage (2) Auteur Éditeur et année d’édition
Page 3 / 5
Course Schedule
Week Titre du Cours Date
1 Part 1: Stochastic Processes and their Classification 24-25 September
Chapter 1 Discret-Time Markov Chains
o Stochastic process
o Markov processes and Markov property
o Discret-Time Markov Chains
o Representative graphs,
o Probabilities and Transition Matrix
2 o States and chains classification 1-2 October
o Stationary distribution
o Examples
3 Chapter 2 Continuous-Time Markov Chains 8-9 October
o Continuous-Time Markov Chains,
o Transition Matrix and representative graphs
o Kolmogorov equations
4 o Stationary distribution and Ergodicity 15-16 October
o Particular Markov processes
Birth and death processes
Poisson Processes
o Some Markovian models
5 Part 2: Introduction to Stochastic Simulation 22-23 October
Chapter 3 Generating of Random Number
o Pseudo random number generator
o Congruential methods
6 o Von Neumann square method 29-30 October
o Validation of pseudo-random numbers
7 Chapter 4 Generating Random Variables for Simulation 5-6 November
Simulation of random variables
o Inverse method
o Accept-Reject method
o Composition method
8 o Specific methods 12-13 November
Binomial variables
Poisson variables
Page 4 / 5
Gaussian variables
9 Simulation of Random Vectors 19-20 November
o Generating multivariate Gaussian vectors
o Generating points in the sphere and in the ball
10 Chapter 5 Monte Carlo Methods and Variance Reduction 26-27 November
Methods
Monte Carlo methods
o Description of the method
o Integration by Monte Carlo method
o Convergence and limit of the method
11 Variance Reduction Methods 5-6 Decemer
o Preferential sampling or importance function
o Control variable
12 o Antithetic variable 12-13 December
o Stratification method
o Average value or conditioning
13 Chapter 6 Simulation of stochastic processes (MCMC) 17-18 December
o Metropolis algorithm
o Generalisation of Metropolis-Hastings
o Boltzman-Gibbs algorithm
14 Part 3: Probabilistic Machine Learning 7-8 January
Chapter 7 Bayesian Networks
Hidden Markov chains
o Bayesian description
o Algorithm of Monte Carlo by Markov chain
15 Approximate Bayesian calculus 14-15 January
Stochastic gradient algorithm and expectation-maximization
Page 5 / 5
