logo-polimi
Loading...
Risorse bibliografiche
Risorsa bibliografica obbligatoria
Risorsa bibliografica facoltativa
Scheda Riassuntiva
Anno Accademico 2019/2020
Scuola Scuola di Ingegneria Industriale e dell'Informazione
Insegnamento 097683 - MACHINE LEARNING
Docente Restelli Marcello
Cfu 5.00 Tipo insegnamento Monodisciplinare

Corso di Studi Codice Piano di Studio preventivamente approvato Da (compreso) A (escluso) Insegnamento
Ing - Civ (Mag.)(ord. 270) - MI (495) GEOINFORMATICS ENGINEERING - INGEGNERIA GEOINFORMATICA*AZZZZ097683 - MACHINE LEARNING
Ing Ind - Inf (Mag.)(ord. 270) - MI (481) COMPUTER SCIENCE AND ENGINEERING - INGEGNERIA INFORMATICA*AP097683 - MACHINE LEARNING
ZZZZ052174 - MACHINE LEARNING (EIT DSC)

Obiettivi dell'insegnamento

Machine learning deals with the problem of building programs that learn from experience. While artificial intelligence focuses on the representation of knowledge and the algorithms for reasoning about such knowledge, machine learning deals with the problem of extracting knowledge from experience gathered solving problems to be used to solve unseen instances of similar problems. 

This course provides the theoretical and methodological foundation of machine learning methods and algorithms that are used in several application areas, including, Data Mining, Robotics, Autonomous Agents, Bioinformatics, Recommender Systems, Social Networks, etc.


Risultati di apprendimento attesi

Dublin Descriptors

Expected learning outcomes

Knowledge and understanding

 

Students will learn how to:

    • Model supervised learning and reinforcement learning problems
    • Train their models
    • Select the model complexity
    • Evaluate the performance of their models

Applying knowledge and understanding

Given specific project cases, students will be able to:

    • Model the learning problem
    • Select the most appropriate learning technique
    • Tune the hyperparameters of the studied techniques
    • Assess the performance of the learned model.

Making judgements

At the end of this course, students will be able to:

    • Correctly assess the performance of the proposed solutions
    • Be aware of fairness issues involved in the training of machine learning methods

Lifelong learning skills 

    • Students will learn the trade-off between bias and variance and hot to manage it by selecting the proper complexity of the learning model
    • Students will learn to develop algorithms to solve relevant learning problems.

Argomenti trattati
  • Introduction (1 class)

Basic concepts. 

  • Learning theory. (4 classes)

Bias/variance tradeoff. Union and Chernoff/Hoeffding bounds.
VC dimension. Worst case (online) learning.
Practical advice on how to use learning algorithms. 

  • Supervised learning. (8 classes)

Supervised learning setup. LMS.
Logistic regression. Perceptron. Exponential family. 
Kernel methods: Radial Basis Networks, Gaussian Processes, and Support Vector Machines.
Model selection and feature selection.
Ensemble methods: Bagging, boosting.
Evaluating and debugging learning algorithms.

  • Reinforcement learning and control. (6 classes)

MDPs. Bellman equations. 
Value iteration and policy iteration. 
TD, SARSA, Q-learning.
Value function approximation. 
Policy search. Reinforce. POMDPs. 
Multi-Armed Bandit.


Prerequisiti

Students are required to know the basics of statistics, linear algebra, calculus, and optimization theory.


Modalità di valutazione

The assessment will be based on a written exam at the end of the course, where both theoretical competence and modeling skills will be tested. 

Type of assessment

Description

Dublin descriptor

Written test

Solution of numerical problems: exercises on supervised learning and reinforcement learning problems

Solution of modeling problems: exercises where the student needs to properly model the machine learning problem and choosing the most appropriate solution technique

Theoretical questions with open answers.

1,2


1,2,3,5


1,2,5

 


Bibliografia
Risorsa bibliografica obbligatoriaChristopher M. Bishop, Pattern Recognition and Machine Learning, Editore: Springer-Verlag Berlin, Heidelberg, Anno edizione: 2006, ISBN: 0387310738
Risorsa bibliografica obbligatoriaRichard S. Sutton and Andrew G. Barto, Reinforcement Learning: An Introduction, Editore: MIT Press, Cambridge, MA, Anno edizione: 2018, ISBN: 0262039249

Forme didattiche
Tipo Forma Didattica Ore di attività svolte in aula
(hh:mm)
Ore di studio autonome
(hh:mm)
Lezione
30:00
45:00
Esercitazione
20:00
30:00
Laboratorio Informatico
0:00
0:00
Laboratorio Sperimentale
0:00
0:00
Laboratorio Di Progetto
0:00
0:00
Totale 50:00 75:00

Informazioni in lingua inglese a supporto dell'internazionalizzazione
Insegnamento erogato in lingua Inglese
Disponibilità di materiale didattico/slides in lingua inglese
Disponibilità di libri di testo/bibliografia in lingua inglese
Possibilità di sostenere l'esame in lingua inglese
Disponibilità di supporto didattico in lingua inglese
schedaincarico v. 1.6.5 / 1.6.5
Area Servizi ICT
30/09/2020