View syllabus
THEORY AND METHODS OF OPTIMIZATION
GIANCARLO BIGI
Academic year2021/22
CourseMATHEMATICS
Code577AA
Credits6
PeriodSemester 1
LanguageItalian

ModulesAreaTypeHoursTeacher(s)
TEORIA E METODI DELL'OTTIMIZZAZIONEMAT/09LEZIONI42
GIANCARLO BIGI unimap
Learning outcomes
Knowledge

The course aims at showing the main theoretical concepts and algorithms for nonlinear optimization problems in finite dimension.

Skills

The course aims at enabling students to formulate, analyse and solve nonlinear optimization problems in finite dimension.

Behaviors

The student should open up the mind to the formulation of concrete problems from different fields as nonlinear optimization problems.

Prerequisites

Linear algebra. Basic notions of topology. Convergence in metric spaces. Multivariate calculus.

Syllabus

Classification of optimization problems. Nonlonear optimization: convex functions and convex sets, local and global minima and maxima,  convex analysis and subdifferential calculus, optimality conditions, duality theory, algorithms for unconstrained optimization (gradient descent, Newton, subgradient, derivate free) and constrained optimization (conditional gradient, projected gradient and subgradient, penalization, interior point, proximal gradient in composite optimization), nonlinear least squares. Equilibria in noncooperative games. Applications to specific problems (e.g., approximation and data/curve fitting, growth models, spatial configurations of moleculse, trasportation on urban and  telecommunication networks, portfolio management, economic measures and their eventual relationships, economic equilibria), optimization in artificial intelligence and machine learning.

Bibliography

No textbook will be adopted. During the classes the instructor will provide a detailed list of references for each topic. Some handwritten lecture notes by the instructor are available as well.

Lecture notes    

http://pages.di.unipi.it/bigi/dida/tmo/lista.html (mainly in Italian)

Main references

  1. J. Nocedal, S.J. Wright, Numerical Optimization, Springer, 1999
  2. M.S. Bazaraa, H.D. Sherali, C.M. Shetty, Nonlinear Programming: Theory and Algorithms, Wiley, 1993
  3. D. Bertsekas, Nonlinear Programming, Athena, 2004
  4. J.-B. Hiriart-Urruty, C. Lemaréchal, Convex Analysis and Minimization Algorithms, Springer, 2006
  5. A. Beck, First-Order Methods in Optimization, SIAM, 2017

For further references check this page

Assessment methods

Students who regularly attended classes (32h or more) are free to choose one of the following examination procedures:

  1. final interview
  2. report and seminar

while those who didn't will be necessarily examed via a final interview on the contents of the course. The seminar will be an approximatively 1h talk on an advanced topic related to the contents of the course, while the report provides a detailed analysis of the same topic. The choice of the topic will be jointly made by the student and the instructor. Once the topic is chosen, the student will have 2 months to deliver the report and give the talk.

Updated: 13/08/2021 13:32