Curriculum
Computer Science for Societal Challenges and Innovation, XXXIX series
Grant sponsor
MUR PNRR DM118
Supervisor

Nicolò Navarin
Co-supervisor
s
TBD
Contact
giovanni.donghi@studenti.unipd.it

Project description
In a context of an ever-evolving data landscape, continual learning emerges as a new, pivotal paradigm, addressing the challenges of adaptability and efficiency in machine learning. It is motivated by the need for models that can learn and adapt over time, acquiring and retaining knowledge from a continuous stream of data. This approach is instrumental in mitigating the costly and energy-intensive process of repeated re-training, as it allows models to adapt to new data and tasks while retaining previously learned knowledge. The essence of continual learning lies in forward transfer, enhancing the learning efficiency by utilizing past knowledge to accelerate the acquisition of new insights. Graph data, characterized by its structured representation of complex relationships and interconnections, is ubiquitous in various domains. From social networks to biological systems, the complexity and dynamic nature of graph data present unique challenges and opportunities for machine learning. I am interested in working on the intersection of continual learning and graph data, a field known as continual graph learning. The goal is to explore and devise solutions that not only mitigate the loss of previously acquired knowledge but also optimize the learning process in the context of dynamic and evolving graph data.