DATA ANALYTICS AND ARTIFICIAL INTELLIGENCE - 1

This course covers part of the "quantitative" aspects of the master program, and aims to provide the student with knowledge and skills on predictive data mining methods (Module 1) and to the founding ideas of Deep Learning (Module 2), offering a broad understanding of both its methodological and practical aspects.

More specifically, the goal of Module 1 is to enable students the understand and exploit predictive data science techniques including both supervised (classification and regression) and unsupervised methods (clustering). The goal of Module 2 is to teach the basic theoretical framework of Neural Networks and to perform a practical, hands-on exploration of Deep Learning. Mathematical notation is paired with quantitative concepts via code snippets in order to help building practical intuition about the core ideas of machine learning and deep learning.

The course includes the adoption of Data Analytics and Deep Learning libraries through the Python programming language.

The course discusses fundamental techniques for predictive and descriptive Data Science and Deep Learning. Students will achieve the following learning outcomes:

Knowledge and understanding:

principles of non-supervised and supervised learning;
principles of data pre-processing and feature engineering;
fundamental aspects of Neural Netowrks and Deep Learning;
details of Neural Network impementation
Applying knowledge and understanding:

being able to apply supervised and unsupervised analysis techniques;
being able to use data analysis software tools (e.g., scikit-learn);
being able to use Deep Learning libraries (PyTorch and Keras).
Making judgements:

being able to choose the most appropriate method to a given problem and to evaluate its performance.

Communication:

reporting comprehensive comparative analysis among different data analysis methods

Students should have achieved the learning outcomes of courses "Computer Programming And Data Management"
(even without passing the corresponding exams).

Module 1:

Introduction to Data Science
Feature engineering: text, numerical and categorical data; importance of similarity functions.
Unsupervised Learning: clustering algorithms, k-means, hierarchical, db-scan; evaluation.
Collaborative filtering: content-based and item-based recommendation algorithms.
Supervised Learning: regression and classification algorithms; logistic classifier, SVM; decision trees; evaluation.
Model tuning and Selection: bias and variance, overfitting, underfitting;
Ensemble methods: Bagging, Boosting, Random Forest.

Module 2:

Neural Networks from scratch
Deep Learning from scratch
Convolutional Neural Networks
Recurrent Neural Networks
Introduction to Keras
Applications with Computer Vision and Text Processing

Module 1: Python Data Science Handbook. Jake VanderPlas. O'Reilly
Module 2: Deep Learning from Scratch: Building with Python from First Principles 1st Edition, Seth Weidman, O'Reilly

This written exam will include three theoretical questions, about topics covered during the whole course, and a small exercise, asking to design from scratch a solution to a practical problem. An example of the exam is provided in the support material section of the Moodle website.

scritto

Il/la docente ha il dovere di vigilare affinché siano rispettate le regole di autenticità e originalità delle prove d'esame. Di conseguenza, nei casi in cui vi sia il sospetto di un comportamento irregolare, l'esame può prevedere un ulteriore approfondimento, contestuale alla prova d'esame, che potrà essere realizzato anche in modalità differente rispetto alle modalità sopra riportate.

Each theoretical question will grant up to 4 points and the exercise will grant up to 20 points.

The students must submit the answers to the theoretical questions within 30 minutes form the start of the exam and the whole solution must be submitted within 90 minutes overall (i.e. if the student submits the theoretical questions in advance, he/she will have more time available for the exercise).

Note that students can obtain up to 2 extra points by submitting the challenges assigned each week by the lecturer and discussed by the teaching assistant during the practice lecture the following week. Specifically 1 extra point will be granted by submitting at least two (fully functional) assignments and 2 extra points will be granted by submitting at least four (fully functional) assignments.

Lessons include both theoretical and hands-on sessions.
Teaching material is delivered through the Moodle platform.
During the course, the python programming language is used together with the scikit-learn library. Students are encouraged to bring their own laptops.