Calendar a.a. 24/25
- Data Science seminars
Laura Sacerdote (University of Turin, Italy)
Location: All lectures will be held in Palazzo Campana (Via Carlo Alberto 10, Torino), Sala Orsi
Timetable: from 16:30 to 19:00 on 16/1; 30/1; 20/2; 27/2; 5/3; 13/3; 20/3; 27/3; 9/4
Contents: The course skeleton is made up of doctoral student seminars. The seminars of the doctoral students aim to increase and motivate the interdisciplinary approach by pushing the doctoral students to a broad vision of data science and to understand its different branches. In addition, the students are motivated to improve their skills in presenting their research topic. Each session includes two seminars of doctoral students while two other students ask them questions. The teacher, together with the students who ask the questions, stimulates the discussion, points out any lack of clarity and suggests further developments of the research. The seminar topics range from computer science to statistics or mathematics based on the research topic of the speaker. The course also aims to foster the emergence of interdisciplinary research collaborations among the doctoral students
- Time series analysis
Elvira Di Nardo (University of Turin, Italy)
Luis Alberiko Gil-Alaña (University of Navarra, Spain)
Location: Palazzo Campana (Via Carlo Alberto 10, Torino) and Corso Unione Sovietica.
Timetable: 31/03, 7/04, 14/04 ,16/04, 23/04, 28/04 (from 11.15), 6/05, 8/05, 13/05, 15/05, 20/05, 22/05, 28/05, 30/05 (from 11.15) for further information contact the teacher
Contents: With this course, the Phd student should be able to transform a real problem into a statistical one and interpret results in an effective way for phenomena evolving during the time. Moreover it is expected that the student is able to employ mathematical/statistical models for a better identification of the dependence and for forecasting the behaviour of the stochastic dynamic system under observation. Syllabus: Strong and weak stationary time series, IID sequence. Time average and almost sure convergence for strong stationary t.s. Mean-ergodic property and stationary time series. Estimation of ACF's functions: statistical properties .Ergodicity covariance property in L^2. Ljung-Box test. Transformations of data and difference operators. Additive models: decomposition in trend and seasonal components and remainder term. Fitting the trend component. Global and local trend: linear trend, regression and approximation. Filtering: two sided moving average, asymmetric filter. Seasonal component: Periodogram. Analysis of the residuals. Wold's decomposition. Linear time series: time invariant linear filter. MA of order infinite. M(q) and AR(p) models. ADF test. Yule-Walker's equations for the covariance function of AR(p) models. ARMA models. Parameter redundancy. Identifiability of a model: causal and invertible representation of an ARMA model. Partial autocorrelation function. Estimation methods for the parameters of ARMA models. the AIC index. Forecasting. Validate the fitting.
-
Complementing big data with small data
Stefano Ferraris (Politecnico di Torino, University of Turin, Italy)
Lorenza Fontana (Politecnico di Torino, University of Turin, Italy)
Location: tba
Timetable: tba (9 hours + 6 hours)
Contents: tba
- Spatial Analysis and Modeling
Rossano Schifanella (University of Turin, Italy)
Location: Computer Science Department, Via Pessinetto 12
Timetable: 4 classes of 4 hours between June-July 2025
Contents: This course explores spatial analysis and geospatial modeling techniques, emphasizing practical applications and the use of modern computational tools. Through theoretical foundations and hands-on practice, students will learn to gather, preprocess, and analyze spatial data while utilizing advanced modeling techniques to address real-world challenges in areas such as urban planning, mobility, and environmental monitoring.
The course is designed to equip students with the knowledge and skills needed to model spatial relationships, identify patterns, and predict outcomes using geospatial data. By the end of the course, students will be able to create, evaluate, and implement spatial models to inform decision-making processes. Course Objectives: Understand the fundamental principles of geospatial data structures and spatial relationships, Master techniques for preprocessing, visualizing, and analyzing spatial data, Develop advanced spatial models for predictive and exploratory purposes.
Apply spatial modeling techniques to real-world scenarios, including urban planning and human mobility. Syllabus:
Foundations of Spatial Analysis: Introduction to Spatial Analysis, Understanding spatial data and relationships. Theoretical principles of spatial analysis. Geospatial Data Fundamentals: Data structures in GIS and map projections. Gathering and preprocessing large-scale geospatial data. Handling spatial datasets with Python and open-source tools. Data Visualization Techniques: Choropleth mapping and other cartographic tools. Web-based mapping technologies and interactive visualizations. Exploratory Spatial Data Analysis (ESDA): Identifying spatial autocorrelation and patterns. Using spatial weights and distance-based relationships. Case Studies in Visualization:
Applications in urban planning, human mobility, and environmental monitoring. Predictive Spatial Models: Developing regression models for spatial data. Spatial clustering. Point patterns. Spatial Network Analysis: Modeling transportation and mobility networks. Measuring connectivity and accessibility within cities.
Hands-On Modeling: Applying advanced modeling tools (e.g., PySAL, GeoPandas) to real datasets.
- Deep learning: an introduction and some mathematical results
Elena Issoglio (University of Turin, Italy)
Location: Palazzo Campana (Via Carlo Alberto 10, Torino)
Timetable: 27/05, 29/05, 30/05, 3/06, 5/06, 6/06 (by Elena Issoglio), 12/06, 13/06 (by guest lecturer Dr A. Ocello), from 10.30am (2 hours) room 2 Palazzo Campana,
Those interested in attending the lectures are kindly invited to let Prof Issoglio know.
Contents: The course is divided into two main parts. Part I offers a (crash) introduction to machine learning, starting from some classical statistical techniques (regression, classification, optimization, etc) and then moving onto the basics of neural networks (some history, architectures, universal approximation theorems, backpropagation algorithm, stochastic gradient descend (SDG), etc). Part II will cover more advanced topics (independent of each other) which are currently topic of research by mathematicians and statisticians in the framework of the mathematical foundations of neural network training. Examples could include central limit theorem-like results for Gaussian neural networks and their explicit rate of convergence to Gaussian processes, score-based generative models and their links to stochastic differential equations, mean-field approximations of NN and link to SDG.