Vai al contenuto principale

PhD Projects (deadline: July, 7 2023 at 12:00 noon Italian time)

The following PhD projects are part of a call for no. 9* (of which no. 1 reserved to candidates holding an international degree) fully funded PhD scholarships in the framework of the PhD in Modeling and Data Science  at the University of Turin (Italy)  - details on the call.

*no. 4  scholarships are funded by the university, no. 1 scholarship is co-funded by ARPA, no. 1 scholarships is funded by LEONARDO group, no. 2 apprenticeship scholarships founded by T.P.S. S.p.A., no. 1 scholarship co-funded by ASI

Please note that further fellowships will be added to this call until the announcement expires (these further fellowships will be positions funded by companies). This web page will be updated as soon as the new positions are approved and published in the annex. 

It is possible to have joint supervision with a partner university (see cotutelle).

The PhD program is interdisciplinary, and it involves branches of mathematics, informatics, economics, statistics, and physics. The scholarship is for three years, starting from 1st November 2024.

How to apply: All interested candidates should submit their application online via the link (the official call for application and further details in english and in italian are published here under the section “Bando/Call - please, see among tags at the top of the page).

The deadline to apply is 20th June 2024 at 12:00 PM (noon) CEST time

Composition of the Examining Boards, dates of entrance examinations and selections results are published here under the section “Selezioni/Selections” (please, see among tags at the top of the page).

Final ranking lists are published here in the section “Graduatorie finali/Final rankings” (please, see among tags at the top of the page).

More information including the official call, all relevant deadlines and FAQs can be found here 

The application includes the submission of a research project from the list below (open the drop-down menu). 

For the applicants: If you have applied for one of the listed titles and a new title of your interest later appears, your application is also eligible for that new project; you can explain your new interest during the interview. However, if you want to change your application by submitting a new research project on the new topic, you can do this by writing to The offices will help you in this change and you will not pay any additional cost.


  • Stochastic processes with applications 

The interested candidate must submit a project based on one of the following subtitles:

1. Stochastic systems for anomalous diffusion. The research will be conducted within the field of stochastic modelling for anomalous diffusion, i.e., random processes suitable to describe systems that exhibit behaviour deviating fundamentally from the simplest diffusion models. The candidate must be ready to work both on purely theoretical aspects (analysis and stochastic processes) and on computational aspects (sampling algorithms). (advisor: Bruno Toaldo)

2. Numerical methods for irregular stochastic differential equations: from standard to deep learning techniques. The research will be conducted within the field of stochastic analysis and their numerical methods. In particular, the candidate will focus on numerical methods for stochastic differential equations with irregular/singular coefficient, both from a theoretical point of view (proof of convergence and convergence rate) and from a practical point of view (implementation of the scheme), comparing results with other techniques (deep learning). (advisor: Elena Issoglio)

  • First passage time problems for stochastic processes with applications

The interested candidate must submit a project based on one of the following subtitles.

1. Approximation of First Passage Time densities. The first-pass time of a stochastic process is the time at which the process first reaches a certain threshold. Characterization of this variable helps to understand the dynamics of the system stochastic evolution under consideration, with remarkable applications in areas ranging from ecology to economics. Because exact solutions can be complex, especially for systems with complex dynamics or high dimensionality, approximations are often pursued, with the aim of making the problem more tractable while still capturing the essential features of the system's behavior. Approximations involve both analytical and numerical and statistical methods. Therefore, the candidate must possess both analytical and numerical/statistical skills. (advisor: Elvira Di Nardo) 

2. Direct and inverse problem for First Passage Time. In several applications the dynamics of the variables of interest is described via suitable stochastic processes constrained by boundaries. Generally the dynamics of the involved variables are described via a suitable diffusion process constrained by an assigned time dependent boundary and the distribution features of the first-passage time (FPT) of the process through time varying boundaries are investigated. This is the direct FPT problem. However, there are also instances when the underlying stochastic process is assigned, one knows or estimates the FPT distribution and wishes to determine the corresponding boundary shape. This is the inverse first-passage time problem. Unfortunately, explicit solutions of the direct or inverse first-passage problems are known only in a limited number of special cases. Hence, the aim of this project is to study the direct and inverse first-passage time problem for stochastic processes from a theoretical, simulative and numerical viewpoint. (advisor: Cristina Zucca)

  • Construction and Bayesian inference for multidimensional diffusion processes (advisor: Matteo Ruggiero)

With potential mobility to one among Duke Univ., USA, UNAM, Mexico, Univ. of Bath, UK

Abstract: The research under my supervision could concern one of the following aspect, to be chosen depending on the background of the student: the construction of stochastic dynamics, typically through urn schemes, interacting particle systems or time-change of other processes, to obtain classes of multidimensional diffusions after appropriate scaling limits; sequential Bayesian inference for some of these classes of models, with the goal  of reconstructing the trajectories and estimating the parameters of the process in a hidden Markov model framework given appropriate data collection schemes. Typical models of interest in my area of expertise find application in population dynamics, mathematical finance and Bayesian inference in a temporal framework.

  • Semantic Aware Communication for Future Wireless Networks (advisor: Valerio Bioglio)

Abstract: The aim of this project is to investigate the idea that including semantic and goal-oriented aspects in future wireless networks can produce a meaningful gain in terms of system effectiveness and sustainability. Going beyond the common Shannon paradigm of correct reception of transmitted packets, semantic communication proposes to take into account also the meaning conveyed by the packets. This concept can be pushed forward to goal-oriented communication, where the impact of correct reception and interpretation of packets is evaluated in terms of the accomplishment of a given task. Combining tools from classical information theory, computer science and semiotics, we propose to re-interpret the state of the art of semantic communications from a telecom point of view.

  • Graph generative models to optimize green infrastructure and sustainable mobility in cities (advisor: Rossano Schifanella)

Abstract: Urban greening interventions and sustainable mobility solutions are increasingly being relied upon to improve the health outcomes and well-being of urban communities and mitigate the environmental footprint of cities. In this project, we will develop large-scale multimodal generative models to provide tools to support fair and equitable management and development of urban nature and sustainable forms of mobility. The goal is threefold: (1) to identify areas that require prioritization for future green interventions according to three classes of indicators, i.e., environmental, structural, and socioeconomic; (2) to simulate the effect of greening interventions by building scenarios that support decision-making processes (3) to simulate the evolution and reconfiguration of the urban street network to promote active mobility.

  • Explainability, privacy and utility in Machine Learning (Advisor: Rosa Meo)

AbstractThe recently approved AI Act by EU Commission obliges companies that implement Machine Learning techniques to explain the outcomes to the users and preserve their privacy. We explore the state of the art of techniques and provide enhancements to better provide at the same time explainability, anonymity and quality in the outcomes from the models.

  • Dialogue Systems, Conversational Interfaces and Natural Language Generation for Artificial Intelligence (Advisor: Luca Anselma/ Alessandro Mazzei)

Abstract: Natural language is the most sophisticated technique that machines can use in order to communicate with humans. Three fields of artificial intelligence that can highly benefit from Dialogue Systems (DS), Conversational Interfaces (CI) and Natural Language Generation (NLG) are (1) Automatic Reasoning, (2) Social Robot Interaction, (3) Assistive Technologies. Therefore, the project will focus on the study, the design and the application of DS, CI and NLG to these fields.

  • Investigation with multi-agent system models, within the framework of the H2020 Green Cities pillar, of urban permeability to butterflies motion and of their diffusion in mixed urban areas (buildings and green spaces). (Advisor: Marco Maggiora)

  • Methods and Tools for Cross Platform Federated Learning (Advisor: Marco Aldinucci)

Abstract: The amount and quality of training data available to single parties often limits the potential of AI. Federated Learning (FL) allows to develop better AI systems leveraging multiparty collaboration without the need to share nor centrally pool the training data. This removed one of the main barriers to collaboration and led to the development of a broad zoo of FL frameworks, each with its own capabilities and peculiarities. The project aims to explore the possibility of collaboration across diverse FL platforms with the intent to simplify FL for users and open up a new business model for data owners, i.e. offering to train models on their datasets in privacy preserving manner.

  • Stochastic control with learning (Advisor: Tiziano De Angelis)

Abstract: Stochastic control is a branch of applied probability that studies how to optimally steer trajectories of (controlled) stochastic processes in order to maximise given reward functions. Its applications span several fields of science, including aerospace engineering, biology, economics, and many others. This project will develop methods for the study of stochastic control problems under partial information, i.e., when the parameters of the model are not fully specified and must be estimated during the optimisation procedure (in a dynamic way). Techniques will range from stochastic calculus to partial differential equations. Both theoretical and numerical aspects of the models will be analysed.

    • Environmental monitoring data and probabilistic evaluation of irrigation water resources in Piemonte  

Abstract: The doctoral project regards the use of environmental monitoring data for the evaluation of the irrigated areas and the water discharges actually utilised in Piemonte. This project aims to quantify areas and discharges in order to prevent agricultural lack of water. Some drought has occurred in the last years, but today, an impressive amount of data is available and a variety of statistical and computer science tools can be applied for the study of this big data. Hundreds of meteorological stations in the Piemonte Region are operated in real time by ARPA every day. Also, the river level monitoring allows us to quantify the amount of water available for diversion in irrigation canals. These data cover more than twenty years and can be compared with satellite data. This project will join specific mathematical models and a machine learning approach to combine ground and satellite data in an effective and unprecedented way. Furthermore, the use of such results will suggest prevention actions in order to limit the drought related risks due also to climate change.  This project allows benefitting of 7 months of scholarship increase for periods abroad (Advisor: Stefano Ferraris).

    • Methods and tools for Cloud-HPC convergence  

Abstract: This research is mainly industrial, with civil use and not classified.  Scholarship co-funded by LEONARDO S.p.A.  (Advisor: Marco Aldinucci).

    • Implementing Artificial Intelligence in Interactive Digital Technical Publishing Services: Innovation and Efficiency (TSP group)  

Abstract: The scholarships are for apprenticeships in higher education and research in the company TPS group for young people aged 18 to 29 years old, aimed at obtaining university and higher education degrees, including PhD. TPS Group is a large company in the field of aerospace and avionics, in Gallarate and Torino. It has a large database of documents relating its business activities, in multiple languages. It wishes to acquire knowledge and competences by means of the acquisition of new forces specialised in innovation technologies, like Artificial Intelligence, Machine learning, natural language processing,  information systems and information retrieval, generative AI, etc. This will allow the company and its customers to improve the management, the knowledge representation and the extraction of information with new services. (Advisor: Rosa Meo)

    • Machine learning applied to remote sensing for land monitoring (ASI)

Abstract: Satellites acquire large volumes of images over the earth lands that can be joined with other open data in order to acquire knowledge about the natural processes (amount of water underneath, presence of fertilisers, amount of CO2 and other gases in the atmosphere, etc.). Extracting knowledge and models from this data will allow to combat and be more resilient to the climate change, will boost the smart agricolture and improve the economy of developing countries. We want to apply multiple Machine Learning technologies (transformers, deep neural networks, graph neural networks) to analyse these data, reconcile them in multiple scales and transfer existing models to fulfil new goals. During the doctoral program the candidate will go 6 months to INRAE (Montpellier, France) and 6 months to Thales (Italy) to work on satellite constellations. (Advisor: Rosa Meo)

Last update: 15/06/2024 06:19
Non cliccare qui!