% IMPORTANT: The following is UTF-8 encoded. This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.
@ARTICLE{Keup:885628,
author = {Keup, Christian and Kühn, Tobias and Dahmen, David and
Helias, Moritz},
title = {{T}ransient chaotic dimensionality expansion by recurrent
networks},
reportid = {FZJ-2020-03969},
pages = {2002.11006},
year = {2020},
abstract = {Neurons communicate with spikes, which are 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 we show that large random networks of rate and binary
units have identical second order statistics. Yet their
stimulus processing properties are radically different: We
discover a chaotic sub-manifold in binary networks that does
not exist in rate models. Its dimensionality increases with
time after stimulus onset and reaches a fixed point
depending on the synaptic coupling strength. Low dimensional
stimuli are transiently expanded into higher-dimensional
representations within this manifold. High noise resilience
persists not only near the edge of chaos, but throughout the
chaotic regime. In rate models of spiking activity, the
effective spiking noise suppresses chaos, severely impairing
classification performance. Chaotic rate networks without
effective spiking noise also show the transient performance
boost. The transitions to chaos in the two models do not
coincide and have qualitatively different causes. Our theory
mechanistically explains these observations. These findings
have several implications. 1) Discrete state networks reach
optimal performance with weaker synapses; implying lower
energetic costs for synaptic transmission. 2) The
classification mechanism is robust to noise, compatible with
fluctuations in biophysical systems. 3) Optimal performance
is reached after only a single activation per participating
neuron; demonstrating event-based computation with short
latencies. 4) The chaotic sub-manifold predicts a transient
increase of variability after stimulus onset. Our results
thus provide a hitherto unknown link between recurrent and
chaotic dynamics of functional networks, neuronal
variability, and dimensionality of neuronal responses.},
cin = {INM-6 / IAS-6 / INM-10},
cid = {I:(DE-Juel1)INM-6-20090406 / I:(DE-Juel1)IAS-6-20130828 /
I:(DE-Juel1)INM-10-20170113},
pnm = {574 - Theory, modelling and simulation (POF3-574) / MSNN -
Theory of multi-scale neuronal networks (HGF-SMHB-2014-2018)
/ neuroIC002 - Recurrence and stochasticity for
neuro-inspired computation (EXS-SF-neuroIC002) / PhD no
Grant - Doktorand ohne besondere Förderung
(PHD-NO-GRANT-20170405)},
pid = {G:(DE-HGF)POF3-574 / G:(DE-Juel1)HGF-SMHB-2014-2018 /
G:(DE-82)EXS-SF-neuroIC002 /
G:(DE-Juel1)PHD-NO-GRANT-20170405},
typ = {PUB:(DE-HGF)25},
url = {https://juser.fz-juelich.de/record/885628},
}