Dr Fabrizio Costa

Fabrizio received his PhD in Computer Science from the University of Firenze, Italy. Before joining Exeter University, he worked at the Universita' degli Studi di Firenze, Italy, at the Katholieke Universiteit Leuven, Belgium and at the Albert Ludwig University of Freiburg, Germany.

Reserach interests:

At a high level, my central research question is: How can we integrate domain expertise into machine learning (ML) systems at the appropriate level of abstraction, extract knowledge from these systems across different layers of abstraction, and design algorithms that can autonomously discover and transition to higher levels of abstraction?To this end, my work focuses on broadening the scope of information that ML algorithms can process by integrating heterogeneous data sources within a unified framework, with attributed graphs serving as the fundamental substrate for representing both structure and features. I also explore ways to expand the range of problems ML can address by studying generative processes for structured information such as graphs, sequences, and relational data, enabling systems not only to capture predictive patterns but also to model the mechanisms that generate them. Finally, I develop algorithms that operate over and discover abstractions, allowing ML systems to autonomously climb the ladder of abstraction from low-level representations to higher-level conceptual structures.

1) The study of input generalization for ML algorithms has lead to the development of efficient and flexible graph kernels and feature generators for graphs (see the EDeN library) with applications in relational learning and chemo- and bioinformatics domains.

2) The study of generalizing ML tasks has lead to the notion of data driven design or Constructive Machine Learning (CML) (see the CML workshops). The objective of CML is the design of algorithms that can build novel objects endowed with desired properties. Such algorithms have a wide spectrum of applications, ranging from drug and protein design to automatic software generation, from music composition to video game assets construction.

Academic profiles:

Google Scholar, DBLP, Research Gate