Dresden 2020 – wissenschaftliches Programm
Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
BP: Fachverband Biologische Physik
BP 40: Systems Biology, Evolution and Neural Networks II
BP 40.4: Vortrag
Freitag, 20. März 2020, 10:15–10:30, ZEU 250
Dynamics, Statistics and Coding in Random Rate and Binary Networks — •Tobias Kühn1,2,3, Christian Keup2,3, David Dahmen2, and Moritz Helias2,3 — 1Laboratoire de Physique Théorique de l'ENS, Paris, France — 2INM-6, Forschungszentrum Jülich, Germany — 3Department of Physics, RWTH Aachen, Germany
Cortical neurons communicate with spikes, discrete events in time. Functional network models often employ rate units that are continuously coupled by analog signals. Is there a benefit of discrete signaling? By a unified mean-field theory for large random networks of rate and binary units, we show that both models can be matched to have identical statistics up to second order. Their stimulus processing properties, however, are radically different: contrary to rate networks [Sompolinsky et al. 1988], the chaos transition in binary networks [van Vreeswijk & Sompolinsky 1998] strongly depends on network size, and we discover a chaotic submanifold in binary networks that does not exist in rate models. Its dimensionality increases with time after stimulus onset and reaches a fixed point that depends on the synaptic coupling strength. Low dimensional stimuli are transiently expanded into higher-dimensional representations that live within the manifold. We find that classification performance typically peaks within a single neuronal time constant, after which performance degrades due to variability in the manifold. During this transient, resilience to noise by far exceeds that of rate models with matched statistics, which are always regular. Our theory mechanistically explains all these observations.