Nuova ricerca

Mauro DELL'AMICO

Professore Ordinario
Dipartimento di Scienze e Metodi dell'Ingegneria

Insegnamento: Automated Decision Making

Ingegneria informatica (Offerta formativa 2023)

Obiettivi formativi

The goal of the course is to present effective and practical tecniquest to solve decision and optimization problems. Tools from heuristic search and mathematical optimization will be studied, analyzed, and applied to a variety of discrete optimization problems such as, the traveling salesman problem, vehicle routing, processor scheduling, packing, etc. Applications of mathematical optimization to the modeling and optimal hyperparameter selection for machine learning techniques, will be also presented.

The course includes laboratory activities using Python

Prerequisiti

Basic knowledge of machine learning, programming in python, and fundation of operations research.

Programma del corso

[2 CFU] Heuristic algorithms. Constructive heuristics, local search, metaheuristics such as Simulated Annealing, Genetic Algorithms, Tabu search, Variable neighborhood Search, Iterated Local Search, Ruin and Recreate.
[2 CFU] Use of a commercial solvers to optimize decision problems represented throw a mathematical model. The Traveling Salesman Problem, models and solution using the iterated subtour elimination addition technique, with static and lazy constraints.
[2 CFU] Use of mathematical models to represent and optimize Machine Learning approaches, Representation of ReLU Networks and construction of adversaria instances.

Metodi didattici

Seminars and laborarotory.

Testi di riferimento

Lecture notes from the teacher

Additional texts
In Pursuit of the Traveling Salesman
Model Building in Mathematical Programming, H. Paul Williams
Local Search A. Aarts, J.K. Lenstra

Verifica dell'apprendimento

There are two possible assesments.

The first one is a written assesment where the student is required to answer to three open questions on the subject presented in the lectures.

The second assesment method is a project work where the student will select an argument and will implement and test solution methods.

The choice of the assesment method is left to the students and can be changed by the same students.

Risultati attesi

Capacity to use heuristic, exact methods and machine learnbing techniques to solve discrete decision and optimization problems.