Ταυτόχρονη προσομοίωση πολλαπλών δικτύων βιολογικών νευρώνων σε παράλληλα συστήματα
Simultaneous simulation of multiple biological neural networks in parallel systems
View/
Keywords
Νευροεπιστήμη ; Νευρώνες ; Προσομοίωση ; Παράλληλη επεξεργασίαAbstract
This thesis focuses on computational neuroscience, aiming to model and simulate neuronal activity
through neural networks. Biological and computational neuron models, such as the Leaky Integrate
and-Fire (LIF), are analyzed to represent the complex dynamics of neural networks. Through the
simulation of these networks, an attempt is made to understand the mechanisms governing brain
function and the computational processes occurring within.
The study includes a comparative analysis of various models and phenomena, such as chimera
states, and applies the LIF model to investigate neuronal behavior in large-scale simulations. To
improve simulation efficiency, parallel processing is employed, leveraging the capabilities of multi
core systems, with the goal of reducing execution time and enabling an extensive analysis of
parameter combinations.
The results demonstrate significant improvements in execution times and speedup, particularly
with a low thread count, while a decrease in performance is observed as threads increase due to
memory bottlenecks. The study concludes that optimizing memory bandwidth usage can enhance
resource efficiency, providing valuable insights into neuronal activity.