Paolo Frasconi
Paolo Frasconi researches neural networks and learning from sequential, structured and relational data. He received his Florence PhD in 1994 and coauthored that year’s paper with Yoshua Bengio and Patrice Simard on learning long-term dependencies. His work includes applications in bioinformatics, including protein structure. He became a full professor in December 2010. In 2017, he coauthored forward- and reverse-mode methods for gradient-based hyperparameter optimization.
- Now at
- University of Florence
- Works
- 307
- Citations
- 16,551
- h-index
- 44
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
1994
Recurrent networks and learning limits
Frasconi received his computer engineering PhD at Florence in 1994, with a thesis on recurrent neural networks and adaptive sequence processing. That year, he coauthored a paper with Yoshua Bengio and Patrice Simard on the difficulty of learning long-term dependencies.
-
December 2010
A Florence professorship
Frasconi became a full professor at Florence in December 2010. He also held earlier Florence appointments and visited other institutions.
-
2017
Optimizing the learning process
Frasconi coauthored a study of forward- and reverse-mode gradients for hyperparameter optimization. The ICML paper computes how validation error changes with learning-algorithm settings. Experiments examined data cleaning and learning interactions between tasks.
-
As of October 2026
Structured data and biology
He works on neural networks, probabilistic models and kernels for structured and relational data. His applications include language, pattern recognition and bioinformatics, including work on protein structure and disulfide connectivity.
Sources and credits
Photo: ai.dinfo.unifi.it; AI Lab - People; creator not stated. License: Not stated by source. Source: ai.dinfo.unifi.it. Cropped to a square and resized. Ask us to remove or replace this photo.