Dottorato in NEUROSCIENZE DEL COMPORTAMENTO

Titolo della tesi: Computational approaches to neuronal dynamics. Analysis and model of neuronal activity through spiking simulations

The brain is a wonderful machine, possessing the remarkable ability to adapt, organize, and process large amounts of information while consuming minimal energy. It can also explore a wide range of dynamic patterns, from highly synchronous states like sleep and deep anesthesia to asynchronous states like wakefulness and cognitive tasks. Sleep plays a crucial role in consolidating memories and integrating new information into existing knowledge. A possible approach to this field relies on the investigation and simulation of spiking neural networks (SNNs) which mimic biological neuron communication protocols and have shown to be capable of efficient learning. This thesis presents a comprehensive examination of the impact of sleep on learning and cognition through a multidisciplinary approach. By developing data analysis tools and data-inspired spiking neural network models, the work provides insight into the effects of sleep-like activity on network structure and synapses. Additionally, the thesis addresses the computational challenges associated with simulating large-scale networks of spiking neurons by implementing a GPU library for fast simulation. Data-constrained models that reproduce spontaneous dynamics recorded in experimental datasets are also developed. Furthermore, it highlights the importance of remaining grounded in biological constraints and constantly evaluating the biological plausibility of implemented models through collaboration with experts in biological dynamics and structure and a thorough understanding of experimental data. Overall, the work demonstrates the potential of spiking neural networks to mimic the brain’s fast and power-efficient learning capabilities and provides a deeper understanding of the role of sleep in the consolidation of memories and the integration of new information into existing knowledge.