Optimization : Nonlinear programming

linma2460  2018-2019  Louvain-la-Neuve

Optimization : Nonlinear programming
5 credits
30.0 h + 22.5 h
Q2
Teacher(s)
Nesterov Yurii;
Language
English
Prerequisites
Basic knowledge of Nonlinear Analysis and Linear Algebra.
The target audience is the students interested in scientific computing, machine learning and optimization in engineering.
Main themes
  • General nonlinear optimization.
  • Smooth and non-smooth convex optimization.
  • Interior-point methods.
Aims

At the end of this learning unit, the student is able to :

1

Learning outcomes:

  • AA1.1, AA1.2, AA1.3
  • AA2.1
  • AA5.2, AA5.3

After this course, the student will be able to :

  1. Estimate the actual complexity of Nonlinear Optimization problems.
  2. Apply lower complexity bounds, which establish the limits of performance of optimization method.
  3. Explain the main principles for constructing the optimal methods for solving different types of minimization problems.
  4. Use the main problem classes (general nonlinear problems, smooth convex problems, nonsmooth convex problems, structural optimization ¿ polynomial-time interior-point methods).
  5. Understand the rate of convergence of the main optimization methods.
  6. Two testing computer projects give a possibility to compare the theoretical conclusions and predictions with real performance of minimization methods

Additional benefits :

  • Training in scientific English
  • Experience in solving difficult nonlinear optimization problems
 

The contribution of this Teaching Unit to the development and command of the skills and learning outcomes of the programme(s) can be accessed at the end of this sheet, in the section entitled “Programmes/courses offering this Teaching Unit”.
Content
  • General problem of nonlinear optimization. Black-box concept. Iterative methods and analytical complexity. Gradient method and Newton method. Local complexity analysis.
  • Convex optimization: convex sets and functions; minimization of differentiable and non-differentiable convex functions; lower complexity bounds; optimal methods.
  • Interior-point methods: notion of self-concordant functions and barriers; path-following methods; structural optimization.
Teaching methods
The course is given in 12-15 lectures. The computer projects are implemented by the students themselves with supporting consultations.
Evaluation methods
In the written exam (in English or French) there are four questions, one for each chapter of the course (up to 5 points for each question). The marks for the exam and the exercises are combined in the final mark.
Online resources
The full syllabus (in English) can be downloaded from the web page of the course.
Bibliography
  • Yu.Nesterov. "Introductory lectures on convex optimization. Basic course", Kluwer 2004
  • P. Polyak, « Introduction in optimization », J. Willey & Sons, 1989
  • Yu. Nesterov, A. Nemirovsky, « Interior-point polynomial algorithms in nonlinear optimization », SIAM, Philadelphia, 1994.
Faculty or entity
MAP


Programmes / formations proposant cette unité d'enseignement (UE)

Title of the programme
Sigle
Credits
Prerequisites
Aims
Master [120] in Data Science Engineering

Master [120] in Mathematical Engineering

Master [120] in data Science: Information technology