Artificial Intelligence methods have achieved remarkable success across domains, yet their effective application to dynamic and structured data remains a major challenge. This thesis investigates advanced AI techniques for continual and robust learning in scenarios where data evolve over time and exhibit complex dependencies. The research explores multiple complementary directions to address the limitations of current models in adaptability and resilience. First, continual learning methods are studied to enable neural networks to learn from sequential streams of data without forgetting past knowledge. A distillation-based approach leveraging Vision Transformers is proposed, where attention representations are transferred between teacher and student models, improving stability across tasks. Additionally, a prompt-learning strategy based on CLIP embeddings is developed to dynamically select task-specific prompts, enhancing performance on the downstream tasks. The second line of research of these thesis embraces federated learning, a distributed setting where structured information naturally arises from the collaboration among clients. A novel defense mechanism against backdoor attacks is introduced, exploiting the spectral properties of local data representations to identify and mitigate malicious participants through data synthesis and alignment. Finally, the thesis investigates adaptive backdoor attacks and corresponding defenses, emphasizing that such vulnerabilities represent a critical threat to industrial processes and infrastructures. Overall, the work contributes to the design of AI models capable of continual adaptation, secure collaboration, and effective exploitation of structural information for real-world and industrial applications.
I metodi di Intelligenza Artificiale hanno raggiunto risultati notevoli in diversi ambiti, ma la loro applicazione efficace a dati dinamici e strutturati rimane una sfida significativa. Questa tesi indaga tecniche avanzate di IA per l’apprendimento continuo e robusto in scenari in cui i dati evolvono nel tempo e presentano complesse dipendenze. La ricerca esplora diverse direzioni complementari per affrontare le limitazioni dei modelli attuali in termini di adattabilità e resilienza. In primo luogo, vengono studiati metodi di apprendimento continuo per consentire alle reti neurali di apprendere da flussi sequenziali di dati senza dimenticare le conoscenze acquisite in precedenza. Viene proposto un approccio basato sulla distillazione che sfrutta i Vision Transformer, in cui le rappresentazioni di attenzione vengono trasferite tra modelli teacher e student, migliorando la stabilità. Inoltre, viene sviluppata una strategia di prompt learning basata sugli embedding del modello CLIP, che seleziona dinamicamente prompt specifici per ciascun task, migliorando le prestazioni. La seconda linea di ricerca della tesi riguarda il federated learning, un contesto distribuito in cui le informazioni strutturate emergono naturalmente dalla collaborazione tra i client. Viene introdotto un nuovo meccanismo di difesa contro gli attacchi backdoor, che sfrutta le proprietà spettrali delle rappresentazioni locali dei dati per identificare e mitigare i partecipanti malevoli attraverso tecniche di sintesi e allineamento dei dati. Infine, la tesi analizza attacchi backdoor adattivi e le relative difese, sottolineando come tali vulnerabilità rappresentino una minaccia critica per i processi e le infrastrutture industriali. Nel complesso, il lavoro contribuisce alla progettazione di modelli di IA capaci di adattamento continuo, collaborazione sicura e sfruttamento efficace delle informazioni strutturali per applicazioni reali e industriali.
Tecniche avanzate di Intelligenza Artificiale per l’apprendimento continuo e robusto su dati strutturati
MENABUE, MARTIN
2026
Abstract
Artificial Intelligence methods have achieved remarkable success across domains, yet their effective application to dynamic and structured data remains a major challenge. This thesis investigates advanced AI techniques for continual and robust learning in scenarios where data evolve over time and exhibit complex dependencies. The research explores multiple complementary directions to address the limitations of current models in adaptability and resilience. First, continual learning methods are studied to enable neural networks to learn from sequential streams of data without forgetting past knowledge. A distillation-based approach leveraging Vision Transformers is proposed, where attention representations are transferred between teacher and student models, improving stability across tasks. Additionally, a prompt-learning strategy based on CLIP embeddings is developed to dynamically select task-specific prompts, enhancing performance on the downstream tasks. The second line of research of these thesis embraces federated learning, a distributed setting where structured information naturally arises from the collaboration among clients. A novel defense mechanism against backdoor attacks is introduced, exploiting the spectral properties of local data representations to identify and mitigate malicious participants through data synthesis and alignment. Finally, the thesis investigates adaptive backdoor attacks and corresponding defenses, emphasizing that such vulnerabilities represent a critical threat to industrial processes and infrastructures. Overall, the work contributes to the design of AI models capable of continual adaptation, secure collaboration, and effective exploitation of structural information for real-world and industrial applications.| File | Dimensione | Formato | |
|---|---|---|---|
|
PhD Thesis - Martin Menabue.pdf
accesso aperto
Licenza:
Tutti i diritti riservati
Dimensione
4.75 MB
Formato
Adobe PDF
|
4.75 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/364691
URN:NBN:IT:UNIMORE-364691