Call for application 42nd cycle

Bando ordinario


Educational goals and objectives

DATA SCIENCE The availability of huge volumes of data, basically characterized by the increasingly extensive and pervasive use of digital technologies, leads the ongoing revolution in many areas of social, economic and industrial reality, posing new challenges to the scientific and technological research in Computer Science, Artificial Intelligence and in several other disciplines, from Physics to Economics, from Medicine to Human Sciences. In this context, the accessibility and processing of large amounts of data, both in centralized and distributed systems, their usage in the design and implementation of complex decision-making models, favors the study and development of autonomous systems in different fields (from mission-critical applications to the studies of natural and social phenomena, the prediction of economical dynamics as well as the diagnostics and planning in medicine or in industrial robotics).

The PhD in Data Science is aimed, in its markedly interdisciplinary nature, to train, at the highest level, experts able to conduct research to understand and master the mathematical, statistical and computer science methodologies of data analysis and processing as well as the underlying technologies supporting applications in a wide variety of of scientific, industrial, economic, medical and social contexts.

The reasons for a new PhD program in Data Science are many and significant. First of all, the offer in Italy of a doctoral training on these topics is still limited, but at the same time it faces an increasing demand for experts in Data Science. This training should be understood with a characterization focused on mathematical-computer science skills, then a strong scientific-technological vocation, properly integrated on established statistical, economic, social and linguistic principles, thus able to give answers to the dynamics of this area of expertise that is very accelerated and strongly interdisciplinary and culturally heterogeneous.

The School of Doctorate represents the indispensable ground for a wide range of students of Computer Science, Computer Engineering, Economics, Physics, Mathematics and not only, at our University of Tor Vergata, in Italy and across different countries. These students are attracted by initiatives in Data Science because in them is clear and dominant the role played by the computational knowledge, the strong scientific character of the educational objectives and the diversified potential applications: from Physics to Economics up to Social Sciences.


Available Positions and Scholarships

University scholarship3University scholarship reserved to foreign applicants0Department scholarship0
Scholarship funded by private research institution0Scholarship funded by private non-research institution0Scholarship funded by public research institution0
Scholarship funded by public non-research institution0PhD position under national research programs2PhD position under EU research programs0
Foreign independent candidates2Employee of private research institution/organisation under a cooperation agreement0Employee of private non-research institution/organisation under a cooperation agreement2
Employee of public research institution/organisation under a cooperation agreement0Employee of public non-research institution/organisation under a cooperation agreement2PhD under higher education and research apprenticeship contract at a private research institution/organisation0
PhD under higher education and research apprenticeship contract at a private non-research institution/organisation0PhD under higher education and research apprenticeship contract at a public research institution/organisation0PhD under higher education and research apprenticeship contract at a public non-research institution/organisation0
No scholarship2

Research topics


Founded by: Fondo Italiano di Scienze Applicate (FISA 2004)

Research topic: Mechanistic Interpretability by-design in Neural-Network-based Large Language Models
Description: Mechanistic Interpretability by-design in Neural-Network-based Large Language Models: this theme is in the context of the MeMo project that aims to build alternative ways to build transparent Large Language Models by explicitly using Associative Memories. Memorization is a fundamental ability of Transformer-based Large Language Models, achieved through learning. In this PhD position, the ideal candidate should work on the paradigm shift proposed by MeMo by helping to design an architecture to memorize text directly, bearing in mind the principle that memorization precedes learning. MeMo is a novel architecture for language modeling that explicitly memorizes sequences of tokens in layered associative memories. By design, MeMo offers transparency and the possibility of model editing, including forgetting texts. We experimented with the MeMo architecture, showing the memorization power of the one-layer and the multi-layer configurations. Ideally, the candidate should have a strong mathematical background and strong programming skills applied to neural networks.
Scholarship type: DOTTORANDO SU PROGRAMMA DI RICERCA NAZIONALE

CUP : E83C25003070005
Principal Investigator: Fabio Massimo Zanzotto
Program: FISA-2024-00249
Project: MeMo - Editable Memorization-based Large Language Models based on Explicit Symbolic Knowledge
UpB: ZanzottoF26MeMo
Lenght: 36
Amount: € 75064,35
Founded by: Fondo Italiano di Scienze Applicate (FISA 2004)

Research topic: Injecting Neuro-symbolic Natural Language Processing Capabilities in Neural-Network-based Large Language Models
Description: Injecting Neuro-symbolic Natural Language Processing Capabilities in Neural-Network-based Large Language Models: this theme is in the context of the MeMo project that aims to build alternative ways to build transparent Large Language Models by explicitly using Associative Memories. Memorization is a fundamental ability of Transformer-based Large Language Models, achieved through learning. In this PhD position, the ideal candidate should work on the paradigm shift proposed by MeMo by helping to design an architecture to memorize text directly, bearing in mind the principle that memorization precedes learning. MeMo is a novel architecture for language modeling that explicitly memorizes sequences of tokens in layered associative memories. By design, MeMo offers transparency and the possibility of model editing, including forgetting texts. We experimented with the MeMo architecture, showing the memorization power of the one-layer and the multi-layer configurations. Ideally, the candidate should have a strong mathematical background and strong programming skills applied to neural networks
Scholarship type: DOTTORANDO SU PROGRAMMA DI RICERCA NAZIONALE

CUP : E83C25003070005
Principal Investigator: Fabio Massimo Zanzotto
Program: FISA-2024-00249
Project: MeMo - Editable Memorization-based Large Language Models based on Explicit Symbolic Knowledge
UpB: ZanzottoF26MeMo
Lenght: 36
Amount: € 75064,35

Self-certification of the academic qualification obtained through Model A:

  • Master’s Degree
  • Single-cycle Degree
  • Specialist Degree
  • Old-system Degree
  • Degree obtained abroad corresponding to EQF Level 7
  • provided that they belong to the following classes: All classes

Admission Procedure

Oral interview The oral examination is designed to gain a more detailed understanding of the study and research activities documented in the candidate’s CV and, consequently, to assess their skills and aptitude for doctoral research.
During the interview, particular attention will also be paid to the research project presented by the candidate. The candidate will have the opportunity to discuss the project’s objectives, proposed work plan and expected outcomes, so that the maturity of the candidate’s proposal can be assessed in relation to their CV.
Finally, for candidates taking the examination in Italian, there will be a dedicated examination session to assess their knowledge of English, through the reading and translation of selected extracts from scientific literature in the field of Data Science.
language ITALIANO
INGLESE
Qualifications assessment The assessment of the documents submitted by the candidate will be carried out in accordance with the Commission’s standard procedures and will not require the candidate’s presence. The mandatory documents specified in the Call for Applications will be subject to a preliminary assessment and, when they were not provided, it may result in the exclusion of the candidate who did not submit any of them. The optional documents will be taken into account in the assessment and in the overall score awarded to the candidate.
does not require the presence of the candidates

contacts and info basili@info.uniroma2.it

Required documentation

§ degree thesis abstract
optional, the file must be uploaded within the call deadline
§ research project
mandatory, the file must be uploaded within the call deadline
§ letter of introduction (by a teacher)
The reference letters must be submitted directly by the referees, not by the candidates, in accordance with the procedures specified in the Call for Applications
optional, the file must be uploaded within the call deadline
§ List of publications
optional, the file must be uploaded within the call deadline
§ letter of motivation (by the candidate)
optional, the file must be uploaded within the call deadline
§ publications (a single PDF file containing the publications)
optional
Sono ammesse potenziali pubblicazioni del candidato (articoli scientifici, o technical report) sino ad un massimo di 3, a scelta libera del candidato. Le pubblicazioni dovranno anche essere presenti nell'elenco completo delle pubblicazioni del candidato, sottomesso in un altro documento (file), the file must be uploaded within the call deadline
§ other document: documenti a scelta del candidato, diversi dal titolo di accesso, e che la commissione potrà non tenere in considerazione
optional, the file must be uploaded within the call deadline
§ Curriculum Vitae
optional, the file must be uploaded within the call deadline
§ Foreign Language Proficiency Certificate
optional, the file must be uploaded within the call deadline

Language Skills

the candidate must know the following languages
ENGLISH

Exam Schedule

Oral interview
daynone
notesnone
timenone
classroomnone
addressnone
publication on notice boardnon indicato
publication on the web siteYes
web sitenone
date of publicationnone
contactsnone

Qualifications assessment
daynone
notesnone
La valutazione titoli non prevede la presenza dei candidati.
publication on notice boardnon indicato
publication on the web siteYes
web sitenone
date of publicationnone
contactsnone

Università degli Studi di Roma "Tor Vergata" - Via Cracovia, 50, 00133 Roma RM