Optimization of Complex Systems - 60h

This course provides an understanding of advanced algorithms to tackle general optimization problems.

Instructor: Andrea Brilli

Term: Spring

Location: Mondays Room 48 Via Eudossiana, 18, Roma / Tuesdays Room A5, Via Ariosto, 25, Roma

Time: Mondays 14:00-16:00 / Tuesdays 15:00-18:00

Course Overview

The main goal of this course is to teach you how to design algorithms for different types of optimization problems. We’ll dig into both the practical and theoretical sides of:

  • Derivative-Free methods
  • Constraint-handling techniques
  • Multiobjective problems

By the end, you should have the critical thinking skills to figure out when and why certain algorithmic approaches work better than others for a given problem.

Throughout the semester, we’ll also bring in guest seminars on various topics to keep things fresh and broaden your perspective.

Background material

Reference Books

  • “Derivative-Free and Blackbox Optimization” by C. Audet and W. Hare Link
  • “Introduction to Derivative-Free Optimization” by A.R. Conn, K. Scheinberg, and L.N. Vicente Link
  • “Multicriteria Optimization” by M. Ehrgott Link
  • “Nonlinear Programming: Sequential Unconstrained Minimization Techniques” by A.V. Fiacco and G.P. McCormick Link

Exam

The exam consists of individual projects where you’ll implement optimization methods and explore theoretical topics. You’ll share your findings with the class through dedicated seminars. No exceptions, everyone presents.

About the seminars: the topics of the seminars are part of the program of the course, I will ask questions during the final seminars.

You can find the topics of the projects here. The projects can be done individually or you can make groups of maximum two people, the depth of the project will change accordingly. Starting from April 20th, there will be a weekly meeting for each student/group. The meeting are scheduled every Tuesday 13:00-14:00, and every Friday 10:30-13:00, in my office (A115, Via Ariosto). I will provide a template to fill in order to organize each week.

Schedule

Week Date Topic Materials
1 Feb 24 Course Introduction

Overview of mathematical optimization, course structure, and expectations.

1 Feb 27 Coordinate Search

Basic derivative-free algorithmic structure with convergence analysis

2 Mar 03 Convergence Analysis

Convergence Analysis of Coordinate descent and globalization strategies

2 Mar 06 Direct Search

Direct Search and introduction to complexity analysis

3 Mar 09 Multiobjective Optimization

Introduction to Multiobjective Optimization

3 Mar 10 Multiobjective Optimization

Objective space structure, convex case, portfolio optimization, scalarization methods

4 Mar 24 Multiobjective Optimization

Pareto Front approximation without scalarization

4 Mar 27 Multiobjective Optimization

Quality of Pareto Front approximations

5 Mar 31 Inequality-constrained Optimization

Introduction and Optimality Conditions

6 Apr 13 Inequality-constrained Optimization

Proof of Fritz-John theorem and introduction to the logarithmic barrier

6 Apr 14 Inequality-constrained Optimization

Sequential Interior methods. Theoretical proof and practical considerations

7 Apr 20 Optimization on Manifolds

Seminar by Diego Scuppa

7 Apr 21 Optimization on Manifolds

Seminar by Diego Scuppa

8 Apr 27 Proximal gradient methods

Seminar by Elisa Trasatti

8 Apr 28 Proximal gradient methods

Seminar by Elisa Trasatti