2022-I Call

Application deadline: April 22 (noon), 2022
Requirements: students enrolled at the second semester of the first year of the MSc track, with 20 CFUs and a GPA of at least 27.5
Application: available soon at this website
Available positions:
  • Computer Science and Engineering: 24


Research topics in Computer Science and Engineering

Proposers: Anna Bernasconi
Topic: SARS-CoV-2 variant hunting tool
SARS-CoV-2 variants (such as Alpha/Delta) are named by 'variant hunters' who use time and Excel files to manually identify new, possibly dangerous, mutations that deviate from common behaviors. In this thesis, we design and implement methods for supporting such scientists in automatically monitoring SARS-CoV-2 sequence data and isolating new variant candidates that emerge from existing variants.
Proposers: Cristiana Bolchini
Topic: POLIMI Expertise Map
Politecnico di Milano has a very rich community of experts on numerous research topics, but it is not easy to map expertise and interests. This thesis wants to explore a method to be able to answer a question such as 'Who is working on _research_topic_?' The goal is to create a map of expertise by analyzing scientific publications, theses, and projects available in the institutional repositories.
Additional team members: Pier Luca Lanzi
Proposers: Giacomo Boracchi
Topic: Deep Learning for Non-Matrix Data
Machine/Deep Learning models are very powerful in interpreting arrays like time series and images. Point clouds (PC) are a set of scattered 3D points, which prevent using traditional DL layers (e.g. convolutions) and rather require new models. The thesis addresses the design of new convolutional layers for PCs as well as models to detect anomalies, explain (saliency maps) and generate (GANs) PCs.
Additional team members: Loris Giulivi, PhD student; Luca Frittoli, PhD student
Proposers: Stefano Ceri
Topic: Conversational agent for data science
Data science often isn't accessible to scientists who lack computational skills. DSBot conversational agent combines NLU and AutoML to provide a platform to build scientific workflows starting from questions in natural language. The student will design and implement methods to improve the current prototype in terms of understanding user's requests and automatically generating data analysis pipelin
Proposers: Chiara Francalanci
Topic: User-centric data models
The thesis will analyze the transition to user-centric data models. It will assess the maturity of solutions supporting data-level transactions in the context of a distributed ownership of data. The general goal is to manage data as a property of their owners throughout the data lifecycle. In future senarios, data can also be sold directly by the owners of data, as if data themselves were NFTs.
Additional team members: Paolo Giacomazzi
Proposers: Franca Garzotto
Topic: Emotion-based Adaptive Conversational Agents
In human-to-human conversation a fundamental role is played by emotions, which are often conveyed by, and interpreted from, the sound characteristics of voice. The thesis purpose is to exploit advances in speech-based emotion detection to create a conversational agent (a dialogue system that interacts with humans in natural language) capable to adapt vocal response according to a user's emotions
Additional team members: Fabio Catania (post-doc)
Proposers: Nicola Gatti
Topic: Algorithmic game theory
Algorithmic game theory studies computational problems related to strategic situations where players are rational, e.g., economic transactions, military settings, routing, and game-play problems. A recent subfield of algorithmic game theory focuses on how information can be disclosed (e.g., by persuasion) and diffused (e.g., by social influence) to induce desired behaviors of users.
Proposers: Daniele Loiacono
Topic: Deep Learning for Medical Imaging Understanding
Medical imaging plays a key role in the diagnostic process as well as in the treatment choice. Unfortunately, the analysis, the segmentation, and the features extraction from medical images is still a very time-consuming process. This research aims at the design and application of deep learning methods to relevant medical imaging processing tasks, such as the organs and tumor segmentation and the
Additional team members: Leonardo Crespi
Proposers: Luca Magri
Topic: 3D vision algorithms for dynamic scenes
One of the long-term goals of Computer Vision is to design algorithms for perceiving and digitalizing the world acquired by imaging sensors. Compelling results have already been obtained for static scenes, next research challenges will be to address the case of dynamic ones. The thesis concerns the design of new algorithms for the reconstruction of 3D scenes composed by multiple rigid objects.
Additional team members: Andrea Porfiri Dal Cin (Phd student)
Proposers: Alberto Marchesi
Topic: Multi-Agent Learning
Recently, machine learning techniques for multi-agent settings received a growing attention. These work by letting the agents interact repeatedly, in order to learn strategies that converge to desirable outcomes (a.k.a. equilibria). The goal of this research topic is to design learning dynamics with provable guarantees in terms of speed of convergence and quality of the reached outcomes.
Proposers: Davide Martinenghi
Topic: Ranking Systems
The simultaneous optimization of different criteria (e.g., attributes of a database) is a ranking problem naturally arising in many scenarios, including recommender systems, machine learning, and meta-search. The aim of this study is to use ranking techniques to discover potentially interesting data and measure relevant properties of the result, such as bias and distribution.
Proposers: Marco Masseroli
Topic: Bioinformatics and Computational Genomics
The Bioinformatics research area focuses on big life science data and information of different types, increasingly available in many heterogeneous sources, and their best management, integration, mining and analysis with machine learning and deep learning techniques for knowledge discovery, to better understand complex biological phenomena and improve clinical stratification and treatments.
Additional team members: Silvia Cascianelli, PhD
Proposers: Antonio Rosario Miele
Topic: Enabling Distributed Intelligence in Edge Devices for Smart Environments
In several applications, IoT devices collect data that are then processed by centralized computing systems. Recently, the increasing capabilities of edge devices have enabled the development of several solutions for local data processing. The project aims at enhancing such solutions with distributed intelligence methods to coordinate local computations and make them more effective and efficient.
Additional team members: Francesco Amigoni
Proposers: Luca Mottola
Topic: The Battery-less Internet of Things
Energy harvesting is redefining the energy constraints of traditional battery-powered IoT devices. However, such sources of energy are generally erratic, causing systems to shut down unpredictably. We devise new software techniques to render IoT software immune to periods of energy unavailability.
Proposers: Gianluca Palermo
Topic: High Performance Virtual Screening for Drug Discovery in Urgent Computing Scenario
The project wants to find solutions to accelerate the virtual screening process in drug discovery. The project considers the urgent computing scenario, similar to COVID-19, where the computation has to rely on HPC resources. Possible research directions are on code optimization strategies, exploitation of heterogeneous resources (e.g. GPUs) or quantum computers, and machine learning techniques.
Proposers: Barbara Pernici
Topic: Knowledge extraction from uncertain scientific data sources
The thesis concerns knowledge extraction from scientific data from multiple sources, using data science techniques from different perspectives: data quality and management, uncertainty, and pattern recognition. Current use cases are chemical data for green combustions and tweets for social media analysis.
Additional team members: Edoardo Ramalli, Carlo Bono
Proposers: Francesco Pierri
Topic: Detecting coordinated inauthentic behaviour on social media
In recent times, manipulation of social media platforms, such as the use of inauthentic accounts to promote agendas and narratives in deceptive ways, has become of global concern. The aim of this thesis is to develop tools to detect so-called astroturfing (fake grassroots) campaigns on Twitter and Facebook.
Proposers: Pietro Pinoli
Topic: Explainable AI approaches for drug repurposing
Finding novel usages for approved drugs is paramount in pharmacological research as it reduces time and cost of developing new therapies. Many ML methods have been used for drug repurposing. However the lack of interpretability of ML methods limits the achievement of their full potential. This thesis aims to develop explainable AI approaches to drug repurposing.
Proposers: Rosario Michael Piro
Topic: Bioinformatics (network medicine and pathway analysis)
Network medicine refers to the study of molecular networks (e.g., signaling networks and metabolic pathways) in the context of human disease. Pathway analysis identifies biological processes which are altered, e.g., in cancer versus control. We will develop a new approach, based on network topology, to identify relevant changes at different scales (from small changes to a pathway-wide deregulation
Additional team members: In collaboration with University of Jena, Germany
Proposers: Marcello Restelli
Topic: Reinforcement Learning
Reinforcement learning is a learning paradigm concerned with learning to control a system so as to maximize a numerical performance measure that expresses a long-term objective. Reinforcement learning is rapidly gaining attention both from academic and industrial communities due to recent successes in a variety of sequential decision-making problems.
Additional team members: Alberto Maria Metelli
Proposers: Manuel Roveri
Topic: Tiny Machine Learning
Technological progress in Internet-of-Things has opened the way to a pervasive presence of distributed intelligent applications in our everyday lives. The research will focus on Tiny Machine Learning for the on-the-device training and recall of machine/deep learning models taking into account the constraints on computation, memory, and energy characterizing the hardware platforms.
Additional team members: Ing. Massimo Pavan
Proposers: Monica Vitali
Topic: An Adaptive Scheduler for Green Applications in a Fog Environment
Sustainability and Energy Efficiency are drivers for the improvement of our society and a necessity for a better future. While Data Centers are getting greener, this is not the case for other computing locations like fog and edge computing, composed of small and inefficient data centers. Considering applications as composed of several microservices interconnected in a workflow and exchanging data,
Proposers: Vittorio Zaccaria
Topic: Security in low power embedded systems.
Low power embedded systems are increasingly characterised by confidentiality and integrity concerns. This research aims at investigating the energy/security tradeoffs and devising tools and methods that can enable them.
Proposers: Davide Zoni
Topic: Hardware design of machine learning for wearables
The deployment of machine learning on wearable devices is limited by three factors: privacy, security, and efficiency. This research aims to design novel hardware platforms to support machine learning at the edge. Such platforms must ensure learning abilities while preserving data sources' privacy, security against implementation attacks, and computation efficiency.