Thesis projects
If you want to request a thesis project in collaboration with the RSTLess group, compile !
Bachelor Thesis
Deep Learning Theory
3
Application of neural networks for the classification of observable patterns in ionospheric plasma
The project is still to be defined more specifically, but it will concern the automatic detection of events of interest within time series. A "trivial" but important enhancement is to eventually use explainability techniques in order to analyse the results and provide some sort of physical interpretation.
Advisor: Fabrizio Silvestri
Co-Advisor: Federico Siciliano
Difficulty: ●●○
Stopping the Noise: A Review of Early Stopping Techniques for Learning with Noisy Labels
The primary goal of this thesis is to study the various approaches utilizing early stopping to address the challenge of learning from noisy labels. By examining commonalities and divergences among these methods, we seek to identify potential areas for intervention and improvement.
Advisor: Fabrizio Silvestri
Co-Advisor: Maria Sofia Bucarelli
Difficulty: ●●○
Green Learning: Enhancing Model Selection with Automated Tracking of Performance and Environmental Impact
ENG: This thesis focuses on improving a library that selects machine learning models based on both performance and environmental impact. The proposed enhancement adds automatic logging of training metrics and carbon footprint, eliminating the need for manual input from users. This upgrade aims to streamline sustainable model selection and foster environmentally conscious AI development with minimal user effort. Moreover, we aim to build a user-friendly UI where the user is prompted to input a dataset. Based on this input, our method GREEN performs inference giving in some parameters and automatically generates runnable PyTorch code or a model. This output is optimized to achieve the desired levels of accuracy and emissions.
ITA: Questa tesi si concentra sul miglioramento di una libreria che seleziona i modelli di Machine Learning in base alle prestazioni e all'impatto ambientale. Il miglioramento proposto aggiunge la registrazione automatica delle metriche di training e del consumo di CO2, eliminando la necessità di input manuali da parte degli utenti. Questo aggiornamento mira a semplificare la selezione dei modelli sostenibili e a promuovere uno sviluppo dell'intelligenza artificiale attento all'ambiente con il minimo sforzo da parte dell'utente. Inoltre, vogliamo costruire un'interfaccia utente facile da usare, in cui all'utente viene richiesto di inserire un set di dati. Sulla base di questo input, il nostro metodo (GREEN) esegue un'inferenza fornendo alcuni parametri e genera automaticamente un codice PyTorch eseguibile o un modello. Questo output è ottimizzato per raggiungere i livelli desiderati di accuratezza e di emissioni.
Advisor: Gabriele Tolomei
Co-Advisor: Filippo Betello and Antonio Purificato and Vittoria Vineis
Difficulty: ●●○
Information Retrieval
2
Cross-Domain Similarity using Optimal Transport
Cross-domain recommendation (CDR) has recently emerged as an effective way to alleviate the cold-start and sparsity issues faced by recommender systems, by transferring information from an auxiliary domain to a target domain to improve recommendations. Studying the similarity between domains is a novel direction in CDR research, poten- tially opening doors for further exploration. In this context, we want to introduce an approach to quantify similarity between a pair of domains using optimal transport and explore how current CDR methods perform with both similar and dissimilar domain combinations.
Advisor: Fabrizio Silvestri
Co-Advisor: Federico Siciiliano or Filippo Betello or Antonio Purificato
Difficulty: ●○○
Lighting Up Recommendations: Enhancing the EasyLightning Library for Scalable Deep Learning in Recommender Systems
This thesis aims to enhance the easylightning library to better support the development of deep learning recommender systems. The project focuses on adding modern recommendation models, improving scalability and modularity, and simplifying usage through cleaner APIs and better documentation.
Advisor: Fabrizio Silvestri
Co-Advisor: Federico Siciliano
Filippo Betello
Difficulty: ●○○
Master thesis
Information Retrieval
3
Are LLMs Fair Recommenders?
Starting from this work [https://arxiv.org/pdf/2305.07609.pdf] (RecSys '23), we propose to further investigate the role of LLMs in Recommendation.
Advisor: Fabrizio Silvestri
Co-Advisor: Federico Siciliano or
Cesare Campagnano
Difficulty: ●●○
Enhancing Reproducibility in Recommender Systems Research: A Systematic Evaluation
Reproducibility is a critical challenge in RecSys research, as many published results rely on specific experimental setups that may not generalize across different implementations. This thesis investigates the reproducibility of the best RecSys papers by introducing new datasets, incorporating additional models, and/or leveraging multiple libraries, including RecBole, Elliott, original implementations, and custom implementations. The study aims to assess the consistency of reported results, identify reproducibility gaps, and provide guidelines for improving transparency in RecSys research.
Advisor: Fabrizio Silvestri
Difficulty: ●●○
Dynamic Loss Switching in Recommender Systems: Analyzing Its Impact on Training and Performance
Most RecSys models are trained using a single loss function, yet different loss functions capture different learning dynamics. This thesis explores the effect of dynamically switching loss functions during training. The study will analyze how this strategy influences model convergence, ranking effectiveness, and generalization. Experiments will be conducted on multiple datasets using state-of-the-art RecSys libraries. Findings will provide insights into whether adaptive loss selection can enhance RecSys training.
Advisor: Fabrizio Silvestri
Difficulty: ●●○
Deep Learning Theory
5
Effects of different types of regularization on the complexity of boundaries
Some work stated that for classification, the proper complexity to regularize is the boundary complexity, rather than the functional complexity of F. Indeed the functional complexity . The goal of classification is to recover the Bayes optimal decision boundary, which divides the input space into non-overlapping regions with respect to labels. Therefore, classification is better to be thought of as estimation of sets in R^d, rather than estimation of functions on R^d. ot be closely connected. The set difference reflects the 0-1 loss much more directly than functional norms on F.
Indeed we can have f in the functional space F that approximates the optimal classifier, \eta, so well that ||f -\eta||_{\infty}< \epsilon, but there is still no guarantee of matching the sign of \eta(x)- 1/2 close to the decision boundary. In this work we aim to study the effect of different types of regularization (l_1, l_2 norm, dropout , batch norm) on the boundary complexity and study how this is have an impact on generalization and robustness of the architecture.
Advisor: Fabrizio Silvestri
Difficulty: ●●●
Effects of noisy labels in neural collapse effect
This work starts from the study of the relative position of embeddings of noisy and clean labels in neural networks that have reached training error zero and are further trained. The goal is to observe that the noisy samples are positioned as far as possible from the other class samples. The proposed approach involves starting from the Neural Collapse paper, training it on a clean dataset, examining the embeddings for clean data, training it on a dataset with noisy samples, examining the embeddings for both clean and noisy data, and running their notebook on a noisy dataset. The distance between the embeddings can be measured by computing the angle between the centroids of the two classes and the two noisy labels, particularly in 2D. The findings of this study can help in identifying robust and reliable deep learning models.
Advisor: Fabrizio Silvestri
Co-Advisor: Maria Sofia Bucarelli & Francesco Caso
Difficulty: ●●●
Approximation properties of CEM
The goal of the thesis is to study which kind of functions can be approximated by concepts embedding models.
Advisor: Fabrizio Silvestri
Difficulty: ●●○
Stopping the Noise: A Review of Early Stopping Techniques for Learning with Noisy Labels
The primary goal of this thesis is to study the various approaches utilizing early stopping to address the challenge of learning from noisy labels. By examining commonalities and divergences among these methods, we seek to identify potential areas for intervention and improvement.
Advisor: Fabrizio Silvestri
Difficulty: ●●○
Enhancing Graph Neural Network Accuracy Using a Dual-Network Approach with NCOD Loss Function
It is well-known and observed in our research that underparameterized networks tend not to fit noise in classification tasks, including Graph Neural Networks (GNNs). To leverage this characteristic, we propose developing a dual-network architecture. The first, underparameterized network identifies reliable samples, while the second network employs the NCOD (Noisy Channel Optimized Distillation) loss function on samples with low prediction scores. This approach aims to enhance accuracy beyond the use of NCOD alone.
Advisor: Fabrizio Silvestri
Difficulty: ●●●
NLP
1
Hackathon Creator
This thesis project aims to develop a customized LLM (Large Language Model) dedicated exclusively to the ideation, design, and management of multidisciplinary hackathons for elementary, high school, and university students and early-career professionals. The goal is to create an automated and highly specialized tool to support the entire development and management process of these events.
Advisor: Fabrizio Silvestri
Difficulty: ●●○
Graph Theory
2
Investigating and Optimizing Pooling Layers in GNNs to Mitigate Gradient Bottlenecks
The pooling layer in Graph Neural Networks (GNNs) often acts as a bottleneck, hindering the flow of gradients and impairing model performance. This issue needs to be addressed to enhance the efficiency and effectiveness of GNNs.
Advisor: Fabrizio Silvestri
Difficulty: ●●●
Dual-Architecture Framework Development:
In our previous research, we observed that training Graph Neural Networks (GNNs) for graph classification under label noise conditions revealed a crucial insight: removing the negative eigenvalues of the learned weight matrix prevents overfitting to noise. However, this approach leads to overall training instability. To address this issue, we propose developing a dual-architecture network to enhance stability and robustness in the presence of label noise.
Advisor: Fabrizio Silvestri
Difficulty: ●●●
Code Translation
4
Automatic Code Translation from Spring Boot to Java Native
The objective is to develop a system that leverages market-leading and/or proprietary generative AI tools to automatically translate Spring Boot code into Java Native, minimizing the need for human intervention. The thesis will be in collaboration with Engineering (www.eng.it)
Advisor: Fabrizio Silvestri
Difficulty: ○○○
Large-Scale Code Translation from Java to Rust
The objective is to utilize market-leading generative AI tools and/or develop proprietary models to create a system capable of automatically translating Java code into Rust, preserving its original semantics and improving efficiency. The thesis will be in collaboration with Engineering (www.eng.it)
Advisor: Fabrizio Silvestri
Difficulty: ○○○
Large-Scale Code Translation from Java to Go
The objective is to design a system that employs market-leading generative AI tools and/or proprietary solutions to reliably and scalably convert Java code to Go, minimizing the need for human intervention. The thesis will be in collaboration with Engineering (www.eng.it)
Advisor: Fabrizio Silvestri
Difficulty: ○○○
UI Test Automation from Functional Analysis
The objective is to develop an AI-powered system capable of analyzing functional specification documents (written in natural or semi-structured language) and automatically generating ready-to-use UI tests. The thesis will be in collaboration with Engineering (www.eng.it)
Advisor: Fabrizio Silvestri
Difficulty: ○○○
Want to print your doc?
This is not the way.
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
CtrlP
) instead.