ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING

Single discipline educational activity

Course Sheet

Course code:

CH0007

Academic Year of enrolment:

2022

Year of supply:

2022/2023

Type of Course:

Characterising

Degree:

Course of 2nd Cycle Degree in COMPUTATIONAL COGNITIVE SCIENCE

Disciplinary Sector:

Electronic and Information Bioengineering

Language:

Italian

Location:

CHIETI

Credits:

Programme Year:

Professor and Collaborators:

GUIDOTTI ROBERTO

Cycle:

Second semester

Hours of classroom activity:

Individual study hours:

136

Prerequisites:

Basics of Mathematical Analysis, Probability, Statistics, Linear Algebra, Algorithms and Python Programming.

Objectives

The goal of the course is to provide knowledge on machine learning and artificial intelligence. The student will have the knowledge about the theoretical aspect of the unsupervised and supervised learning, with a particular emphasis on deep neural networks.

The student will be able to apply machine learning algorithms to real-world tasks, to correctly evaluate the proper algorithm to use given a dataset, evaluate the algorithm performances and show the results of the analysis in a clear, effective and critical manner.

A goal of the course is to provide practical tools to easily apply and implement machine learning algorithms.

Contents

The course covers machine learning and artificial intelligence data analysis techniques. During the course, unsupervised and supervised learning will be presented, with a particular emphasis to the deep neural networks. In addition, reinforcement learning will be outlined.

The course will also cover the tools and the methods for the assessment of performances of these algorithms.

During the course will be presented some machine learning libraries, written in Python programming language, that implement the introduced algorithms and that can be used to easily implement new algorithms, as well. These concepts will be taught using practical exercise to ease the comprehension and the usage of the libraries.

Extended Syllabus

Artificial Intelligence and Machine Learning Introduction:
1) Historical outlines
2) Introduction to learning techniques

Unsupervised Learning:
1) Clustering algorithms, k-means, hierarchical clustering, DBSCAN
2) Expected Minimization, Gaussian mixture models, Unsupervised Hidden Markov Models
3) Manifold learning, Multidimensional Scaling, Principal Components Analysis (PCA), Independent Component Analysis (ICA).

Supervised Learning:
1) Classification techniques: Logistic Regression, Gaussian Naive Bayes, SVM, Kernel SVM, Nearest Neighbor.
2) Regression techniques: Linear Regression, Ordinary Least Square, Regularization techniques.
3) Model selection: Cross-validation, Feature Selection, Performance assessment.

Neural Networks
1) Classical Neural Networks: perceptron, feedforward neural networks, backpropagation.
2) Deep Neural networks: Convolutional Neural Networks, Recurrent Neural Networks, Autoencoders, Generative Adversarial Networks.

Introduction to Reinforcement Learning.
1) Markov Decision Process, Bellman equations, Q-Learning.
2) Applications of Deep Reinforcement Learning.

Recommended Bibliography

Textbooks used in the course are:

1) Pattern Recognition and Machine Learning - C. Bishop. Springer, 2006.

2) The Elements of Statistical Learning - T. Hastie, J. H. Friedmann, R. Tibshirani. Springer. 2009. In lingua inglese e disponibile online.

3) Deep Learning - I. Goodfellow, Y. Bengio, A. Courville. MIT Press, 2016.

4)Artificial Intelligence: A modern approach - S. J. Russel, P. Norvig. Prentice Hall, 4th Edition.

Further material for exam preparation and project assignment will be provided by the lecturer on the e-learning platform.

Methods of Provision

Conventional

Teaching Methods

The course consists of 64 hours of frontal teaching, divided into lessons of 2 and 3 hours.

The lectures take advantage of the support of slides and concern theoretical aspects of the discipline.

The course includes practical exercises which concern the application and coding of the algorithms presented during the course. The practical exercises are based on python programming language and python data analysis and machine learning libraries (e.g. scikit-learn, numpy, pandas, pytorch).

Attendance is optional but highly recommended.

Evaluation methods

Verification of learning:

The verification of learning consists in two parts: a written test and a project.

The written examination aims at assessing the understanding of technical and theoretical aspects of machine learning and artificial intelligence presented during the course.

During the project assignment, the candidate applies the topics studied during the course. The project can be developed in teams, with a recommended size of 3 students each group. The project is periodically evaluated, during these sessions the team will show the proposal, the implementation and the results of the project. Finally, the team will present a report that illustrates the activities performed during the development of the project. The projects are agreed with the lecturer.

The exams can be taken with no predefined order.

The final mark, out of thirty, takes into account both tests: the mark of the project (up to 8 points) and the mark of the written test (up to 22 points). The exam is considered passed when the candidate obtains at least 18 points, summed over the tests.

Contacts/More Information

Slides of the course and other teaching material are available on the e-learning platform of the course.

Full frequency to lessons is highly recommended.

Università degli Studi
"G. d'Annunzio"

Chieti - Pescara

Aree

Error message