Ateneo - UPO

The cours is structured in two different logical modules: Machine Learning (6 CFU) and Deep Learning (3 CFU)
MACHINE LEARNING
Supervised and unsupervised models of machine learning;
Regression and classification;
Bias and variance, regularization;
Clustering methods;
Anomaly detection;
Evaluation of machine learning models;
Machine learning tools.
DEEP LEARNING
Theory and Implementation of Deep Learning methods

A. Geron: Hands-on Machine Learning with scikit-learn and Tensorflow,(3rd edition), O’Reilly, 2019;
P. Flach, Machine Learning, Cambridge Univ. Press, 2012;
I.H. Witten, E. Frank, M.A. Hall, C.J. Pal. Data Mining: Practical Machine Learning Tools (4th ed.), Morgan Kaufmann, 2016.
I. Goodfellow, Y. Bengio, A. Courville. Deep Learning, MIT Press, 2016

Knowledge and understanding:
the class will introduce the basic notions of machine learning methods, how to evaluate them and how to develop the related systems.
Applying knowledge and understanding:
the acquired knowledge will allow the student:
to know the concepts of bias and variance of the models and how to deal with the related problems (e.g. regularization techniques)
to know the main classification, regression and unsupervised models most used in current applications
gain an understanding of deep learning that can be applied to a wide range of topics with particular focus on computer vision, word and text processing, signal processing applications.
Making judgements: contents in this class will allow the student to develop autonomous judgement skills; he/she will be able to evaluate different choices concerning the possible machine learning models.
Communications skills: the student will develop relational and communication skills that will allow him/her to work in team, and to emphasize pros and cons of the possible machine learning models.

Learning skills: the study of the principles of the discipline will allow the student to have a correct understanding of the possible alternatives in the development of a machine learning system, and it will allow him/her to choose the most suitable approach.

Probability and statistics, Calculus, Discrete mathematics (matrix calculus), Python language (base)

Classroom lectures and development of simple prototypes with practicals.

Lectures material to complement suggested books provided through DIR teaching platform.
Availability of recorded lectures.

Oral examination with discussion of a simple project.

MACHINE LEARNING
Introduction to machine learning: supervised and unsupervised learning;
Linear regression (univariate and multivariate), ridge regression;
Logistic regression;
Regularization;
Neural Networks;
ML evaluation;
Decision Trees;
Bayesian Learning;
Lazy Learning;
Ensemble learning methods;
Clustering;
Dimensionality reduction: PCA;
The Weka tool;
The scikit-learn APIs.
DEEP LEARNING
Deep Learning theory
Deep feedforward neural nets
Convolutional networks
Recurrent nets
Autoencoder
Attention e Transformers
Pre-Training
Applications
Implementation Packages

KNOWLEDGE:
to know the main methodologies of machine learning (supervised and unsupervised) and of deep learning
to know methods and techniques for the evaluation of the models
to know the suitability features of ML models with respect to a given application
COMPETENCE AND SKILL
to be able to develop an ML/DL model relative to a specific application
to be able suitably deal with bias and variance of the models
to be able to evaluate the results of an ML system, following suitable metrics
to be able to use and apply some of the main tools for ML/DL systems development.
to be able to connect notions presented in different parts of the class; to be able to suitably interface with application experts; to know the right terminology.