Alberto Testolin

Personal Page of Alberto Testolin

Alberto Testolin



Department of General Psychology,
University of Padova
via Venezia 12/2
35131 Padova (Italy)

tel : +39 049 827 6528
fax: +39 049 827 6600
e-mail: alberto.testolin(at)
office: room P13 (ground floor), Building Psico2


Research Interests

I'm interested in artificial intelligence, machine learning and cognitive neuroscience. I study how information is represented and processed in neural systems by means of computational simulations, using deep learning, recurrent neural networks, and related types of probabilistic graphical models.

I started my research by investigating visual word recognition and sequence learning, in order to better understand how the brain might process written patterns and orthographic structures. I am also interested in investigating how cultural artifacts are influenced by - and influence - the human brain. In particular, I am fascinated by the neuronal recycling hypothesis, which suggests that phylogenetically older cortical circuits might be used to accommodate the development of new cultural inventions, such as reading and arithmetic. More recently, I started to explore the predictive coding theory, which sees the cerebral cortex as a generative model of the environment that continuously - and actively - tries to anticipate the flow of sensory information. My current research goal is to exploit this theoretical framework to describe attentional mechanisms from a computational perspective.
I exploit parallel computing architectures to reduce the computational time required to train large-scale neural networks, using both graphic processors (GPUs) and multi-core clusters.

I am a member of the IEEE Task force on Deep Learning.

Ph.D., Cognitive Science, University of Padova (2015)
Laurea (M.Sc.), Computer Science (Artificial Intelligence), University of Padova (2011)

Representative publications (see full list and PDF files here)

  • Zorzi M, and Testolin A (2018). An emergentist perspective on the origin of number sense. Philosophical Transactions of the Royal Society B: Biological Sciences.
  • Testolin A, Stoianov I, and Zorzi M (2017). Letter perception emerges from unsupervised deep learning and recycling of natural image features. Nature Human Behaviour.
  • Testolin A, De Filippo De Grazia M, and Zorzi M (2017). The role of architectural and learning constraints in neural network models: A case study on visual space coding. Frontiers in Computational Neuroscience.
  • Testolin, A., and Zorzi, M. (2016). Probabilistic models and generative neural networks: towards a unified framework for modeling normal and impaired neurocognitive functions. Frontiers in Computational Neuroscience.
  • Testolin, A., Stoianov, I., Sperduti, A., and Zorzi, M. (2015). Learning orthographic structure with generative neural networks. Cognitive Science.
  • Pasa, L., Testolin, A., and Sperduti, A. (2015). Neural networks for sequential data: a pre-training approach based on Hidden Markov Models. Neurocomputing.
  • Testolin, A., Stoianov, I., De Filippo De Grazia, M., & Zorzi, M. (2013). Deep unsupervised learning on a desktop PC : A primer for cognitive scientists. Frontiers in Psychology.