MARIA SOFIA BUCARELLI

Dottoressa di ricerca

ciclo: XXXVI


supervisore: Fabrizio Silvestri

Titolo della tesi: Improving Reliability in Deep Learning: Exploring Generalization Bounds, Noisy Label Handling, and Loss Surface Characterization

This dissertation investigates the theory of generalization and robustness in deep learning. Through diverse research works, the thesis provides valuable insights and advancements towards building more reliable systems. We focus on handling noisy labels, deriving generalization bounds for clustering, enhancing interpretability, and characterizing the topology of the loss landscape. These findings contribute to the broader field of deep learning, advancing the development of effective and reliable machine learning systems. After the introduction provided in Chapter 1, in the second chapter we tackle the challenge of noisy labels in classification by leveraging inter-rater agreement and estimating the noise distribution, thereby improving model performance and robustness. In the third chapter, we establish generalization bounds for projective clustering, presenting near-optimal results for subspace clustering. Chapter 4 introduces a novel artificial neuron that enhances interpretability while retaining the representation power and performance of a standard neural network. Indeed, we prove the universal approximation theorem for specialized versions of the artificial neuron. In Chapter 5, we characterise the topological complexity loss surfaces using Betti Numbers. Understanding the topology of loss surfaces is crucial for studying generalization and robustness in deep learning models. The last chapter summarizes the key findings and contributions, also mentioning possible future directions. Collectively, this research work contributes to understanding generalization and robustness in deep learning, advancing the field and enabling the development of more reliable models.

Produzione scientifica

11573/1717634 - 2024 - A topological description of loss surfaces based on Betti Numbers
Bucarelli, Maria Sofia; D'inverno, Giuseppe Alessio; Bianchini, Monica; Scarselli, Franco; Silvestri, Fabrizio - 01a Articolo in rivista
rivista: NEURAL NETWORKS (Elsevier Science Limited:Oxford Fulfillment Center, PO Box 800, Kidlington Oxford OX5 1DX United Kingdom:011 44 1865 843000, 011 44 1865 843699, EMAIL: asianfo@elsevier.com, tcb@elsevier.co.UK, INTERNET: http://www.elsevier.com, http://www.elsevier.com/locate/shpsa/, Fax: 011 44 1865 843010) pp. - - issn: 0893-6080 - wos: WOS:001262458000001 (0) - scopus: 2-s2.0-85196948612 (0)

11573/1673795 - 2023 - False Data Injection Impact on High RES Power Systems with Centralized Voltage Regulation Architecture
Bragatto, Tommaso; Bucarelli, Marco Antonio; Bucarelli, Maria Sofia; Carere, Federico; Geri, Alberto; Maccioni, Marco - 01a Articolo in rivista
rivista: SENSORS (Basel : Molecular Diversity Preservation International (MDPI), 2001-) pp. 1-17 - issn: 1424-8220 - wos: WOS:000946895200001 (2) - scopus: 2-s2.0-85149799109 (7)

11573/1685080 - 2023 - Leveraging Inter-Rater Agreement for Classification in the Presence of Noisy Labels
Bucarelli, Maria Sofia; Cassano, Lucas; Siciliano, Federico; Mantrach, Amin; Silvestri, Fabrizio - 04b Atto di convegno in volume
congresso: IEEE Conference on Computer Vision and Pattern Recognition (Vancouver; Canada)
libro: IEEE Conference on Computer Vision and Pattern Recognition - (979-8-3503-0129-8)

11573/1707394 - 2023 - On Generalization Bounds for Projective Clustering
Bucarelli, Maria Sofia; Larsen, Matilde; Schwiegelshohn, Chris; Toftrup, Mads - 04b Atto di convegno in volume
congresso: Advances in Neural Information Processing Systems (New Orleans; USA)
libro: Advances in Neural Information Processing Systems 36 (NeurIPS 2023) - ()

11573/1657031 - 2022 - NEWRON: A New Generalization of the Artificial Neuron to Enhance the Interpretability of Neural Networks
Siciliano, Federico; Bucarelli, Maria Sofia; Tolomei, Gabriele; Silvestri, Fabrizio - 04b Atto di convegno in volume
congresso: IEEE International Joint Conference on Neural Networks (Padova; Italia)
libro: IEEE International Joint Conference on Neural Networks - (978-1-7281-8671-9)

© Università degli Studi di Roma "La Sapienza" - Piazzale Aldo Moro 5, 00185 Roma