Speaker
Arvind Kumar
(KTH)
Description
Spatial and temporal patterns of neuronal activity in the
brain are thought to underlie every meaningful behavior.
Consistent with this idea, spatio-temporal sequences have
been observed in different conditions across the brain in
the form of travelling waves and spike patterns. Three
somewhat related mechanisms have been proposed to explain
the existence of temporal sequences. First, it is assumed
that the feedforward networks are embedded in an otherwise
random network. Second, temporal sequences reflect a
systematic transition of the network activity from one
attractor state to another. Such transitions are governed by
specific connectivity rules or neuron and synapse
properties. Third, randomly connected networks learn to
generate sequences using a supervised learning algorithm.
All these mechanisms are untenable given the known anatomy
of the brain and the biological implausibility of supervised
learning. Thus, despite the ubiquity of sequential activity
of neurons, the underlying mechanisms have remained obscure.
To better understand the spatio-temporal patterns in
neuronal activity we investigated the dynamics of neuronal
network with spatial connectivity. We found that to form
spatial pattern in the network activity, spatial
connectivity should vary non-monotonically as a function of
distance between neurons. On the other hand when the spatial
connectivity is asymmetric and inhomogeneous in the network
space, the spatial patterns become unstable and result in
the emergence of temporal patterns. Finally, we derive the
constraints on the asymmetry and inhomogeneity of the
connectivity that results in spatio-temporal sequences
resembling the sequential neuronal activity in the brain.
