Il corso intende fornire una panoramica sugli strumenti principali utilizzati per l'analisi di grandi moli di dati (audio, video, testo) generati dagli odierni sistemi di telecomunicazione e dai relativi servizi offerti.
A tale scopo sono introdotti i principi di inferenza statistica e di machine learning, oltre che le principali architetture di deep learning utilizzati ad oggi in svariati ambiti applicativi.
Sono previste esercitazioni in Matlab relative agli argomenti esposti, e al suo uso per calcolo parallelo.
Il programma di dettaglio del corso prevede:
Statistics
inference and statistical hypothesis testing
regression
Machine Learning
classification (supervised learning)
decision trees, random forests, naïve Bayes, linear discriminant analysis, k-nearest neighbor, support vector machines
clustering (unsupervised learning)
k-means clustering
hierarchical clustering
data modeling
principal component analysis, indipendent component analysis, outlier detection and data cleansing, hidden Markov models
deep learning & CNN
Processing
parallel processing
examples in Matlab
Data analytics in business applications
Graph-based signal processing (TBD)
Students' presentations
Course Contents
The course aims at providing an overview of the main tools used for the analysis of big data (audio, video, text) generated by today's telecommunications systems and related offered services.
For this purpose, the principles of statistical inference and machine learning are introduced, as well as the fundamental deep learning architectures used nowadays in several application areas.
Matlab exercises related to the presented arguments, and to its use for parallel computation, are planned.
The course detail program includes:
Statistics
inference and statistical hypothesis testing
regression
Machine Learning
classification (supervised learning)
decision trees, random forests, naïve Bayes, linear discriminant analysis, k-nearest neighbor, support vector machines
clustering (unsupervised learning)
k-means clustering
hierarchical clustering
data modeling
principal component analysis, indipendent component analysis, outlier detection and data cleansing, hidden Markov models
