Theoretical and Computational Neuroscience
Author: Gabriel Gregorio Torre | Email: gabriel0torre@gmail.com
Gabriel Torre1°2°, Sergio Lew1°3°
1° Universidad de Buenos Aires, Instituto de Ingeniería Biomédica, Buenos Aires, Argentina
2° Universidad de San Andrés, Laboratorio de Inteligencia Artificial y Robótica, Buenos Aires, Argentina
3° Instituto de Biología y Medicina Experimental, CONICET, Buenos Aires, Argentina
In this poster, we investigate an excitatory-inhibitory (E-I) network composed of 1200 leaky integrate-and-fire (LIF) neurons, organized into fully connected clusters: one inhibitory group, multiple selective groups, and one non-selective group. Building on the framework established by Lew and Tseng (2014), we examine how dopamine modulates GABAergic circuits. We explore the monotonic relationship between dopamine levels and oscillation frequency within the network, observing that high dopamine levels disrupt oscillatory patterns, while low levels induce chaotic behavior. We hypothesize that this chaotic activity can be harnessed for state space exploration, facilitating effective sampling in reinforcement learning. This approach, akin to simulated annealing, begins with high randomness and aims to stabilize into desired orbits as reinforcement gradually guides the network. This framework suggests a novel method for leveraging chaotic dynamics in neural circuits to enhance learning and adaptability.