Curriculum
Computer Science for Societal Challenges and Innovation, XLI series
Grant sponsor
Bruno Kessler Foundation
Supervisor
Alessandro Sperduti
Co-supervisor
to be defined
Contact
gabriele.fiaschi@phd.unipd.it
Project description
The PhD project, Neuroscience and AI, proposes an innovative approach to bridge the gap between artificial intelligence (AI) and computational neuroscience. Traditional artificial models, though inspired by neuroscience, often fail to reproduce the temporal behavior of biological neurons and lack the structural complexity of the actual brain organization. The project is based on the Shallow Brain Hypothesis, which suggests reducing the number of layers to two and modifying computational units with recurrent loops and parallel connections to simulate interactions between different brain areas. This methodology offers significant advantages in learning speed and decision-making rapidity compared to deep networks. The core idea is to model the entire neural system by developing a Whole Brain Model, while specifically focusing on the collective models derived from the Shallow Brain Hypothesis. This leads to the use of Reservoir Computing (RC). RC is an approach that resembles the dynamic response of neuronal clusters, offering greater flexibility, improved learning speed, and the capacity to extract information from complex inputs. The main objective is to employ Reservoir Computing models within the Shallow Brain Hypothesis framework. This will be able to avoid the backpropagation algorithm, improving the speed and decision-making process. The project is highly interdisciplinary, drawing inspiration from computer science, neuroscience, mathematics, and physics. The central expected outcome is the development of a neural network that, after training on fMRI and PET datasets such as ADNI and TCIA, will be able to recognize Alzheimer's disease and tumoral clusters. This capability is expected to support earlier diagnosis and ultimately contribute to better patient outcomes.