Berlin 2024 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 18: Pattern Formation, Delay and Nonlinear Stochastic Systems
DY 18.5: Vortrag
Dienstag, 19. März 2024, 10:45–11:00, BH-N 128
Detecting a periodic signal by a population of spiking neurons in the weakly nonlinear response regime — •Maria Schlungbaum1,2 and Benjamin Lindner1,2 — 1Physics Department, Humboldt University Berlin — 2Bernstein Center for Computational Neuroscience Berlin
Signal detection is a ubiquitous problem in several situations for living organisms. We are specifically interested in detecting a weak signal in the presence of a stronger stimulus and noise -- known as cocktail party problem in auditory perception. We simplify this problem here and study the response to two periodic signals by a homogeneous population of stochastic leaky integrate-and-fire (LIF) neurons. Using a threshold-crossing of the population activity as a detection criterion, we show by the means of the receiver operating characteristics (ROC), that the detectability depends strongly on the stimulus amplitude but only weakly on the time window of observation. Interestingly, the detection of a weak periodic signal can be boosted by a strong periodic stimulus. This effect depends on the frequencies of the two signals and the dynamical regime in which the neurons operate. We also present an analytical approximation for the ROC curve based on the weakly nonlinear response theory for a stochastic LIF model.
Keywords: signal detection; nonlinear response theory; integrate-and-fire models; receiver operating characteristics