Teaching AI to learn faster
Most artificial intelligence (AI) models many examples to be able to learn new things, which can be problematic when there is a lack of data. However, the results of Anna Vettoruzzo’s doctoral studies could change the way AI is used in areas where data collection is expensive or impractical, such as healthcare and autonomous systems.
In her doctoral thesis, Anna Vettoruzzo explores how meta-learning can help AI models learn faster by using what they already know. Meta-learning is sometimes called “learning to learn”, as it is a kind of machine learning that trains AI models to learn how to learn and adapt to new tasks on their own.
Anna Vettoruzzo defended her PhD thesis in February 2025.
Anna Vettoruzzo’s research explores how meta-learning can be integrated with other fields of machine learning to improve both efficiency and generalisation, which is the ability to apply what the model has learned from one task to a new, similar task. The results of Anna Vettoruzzo’s studies show that meta-learning is particularly effective in handling scenarios with high task variability, such as helping an AI model adapt to different areas (domain adaptation), continuously learning from new information without forgetting (continual learning), training without sharing private data (federated learning) and training with unlabelled data (unsupervised learning).
More adaptable AI systems
Throughout her research, Anna Vettoruzzo has investigated how meta-learning can be integrated into diverse machine learning fields to improve model generalisation through efficient adaptation to new tasks with minimal supervision. By using meta-learning to efficiently adapt to new tasks, AI models can get closer to the human way of learning new concepts.
“This discovery highlights new possibilities for building more adaptable AI systems that require minimal supervision while maintaining strong generalisation capabilities. It was surprising to see how meta-learning can be applied to different machine learning fields to enhance not only the performance of the models but also their generalisation ability with only limited amounts of labelled data”, Anna Vettoruzzo explains.
Healthcare is an area where data collection is impractical and meta-learning can be useful.
Accessible AI in real-world industries
The results of Anna Vettoruzzo’s research could change the way AI is used in areas where data collection and annotation are expensive or impractical, such as healthcare, personalised education and autonomous systems.
“By improving the generalisation capabilities of machine learning models, my work helps create more flexible and efficient AI systems that can operate in dynamic environments, ultimately reducing the need for human intervention and making AI more accessible and scalable across various domains”, Anna Vettoruzzo says.
Anna Vettoruzzo’s educational background
Anna Vettoruzzo received her Bachelor of Science degree in Information Engineering and her Master of Science degree in Information and Communications for Internet and Multimedia with a focus on Machine Learning for Healthcare at the University of Padova, Italy. She recently finished her doctoral studies in Information Technology at Halmstad University with a thesis called “Advancing Meta-Learning for Enhanced Generalization Across Diverse Tasks”.
Anna Vettoruzzo chose to do her PhD in Sweden because of its strong research environment, and Halmstad University because of a research project about making machine learning systems better at mimicking how humans learn. Throughout her PhD journey, she has collaborated with researchers from both Halmstad University and other universities, such as Eindhoven University of Technology, the Netherlands.
In the immediate future, Anna Vettoruzzo will be joining Eindhoven University of Technology as a postdoctoral researcher, where she will apply meta-learning for designing efficient fine-tuning large language model (LLM) techniques. Looking ahead, her long-term goal is to become a full professor, allowing her to both conduct impactful research and mentor the next generation of scientists.
Text: Emma Swahn
Photo: Magnus Karlsson, Pixabay
More information
Research at the School of Information Technology
Read Anna Vettoruzzo’s thesis: Advancing Meta-Learning for Enhanced Generalization Across Diverse Tasks
External link.
