Thomas Rost, FU Berlin

Modelling Cortical Variability Dynamics - From Inhibitory Clustering to Context Dependent Modulation

The variability statistics of cortical spike trains are analysed in the framework of renewal process theory. The estimation of interval variability is refined so that it can be used to extract the underlying firing rate variance from neural activity. These methods are then applied to a cortical data set where we show that the reduction in count variability commonly observed at the onset of stimuli can be dissected into a modulated component of rate variance and an interval variability that remains constant. The reduction of count variability is commonly modelled using balanced networks of excitatory and inhibitory units where the excitatory population is divided into clusters of stronger internal connectivity.
Using a mean field description of balanced networks of binary neurons, we show that the firing rates of the active clusters are regular and often close to the saturation limit. As a possible remedy we propose to subdivide the inhibitory population into clusters as well so that each excitatory cluster is selectively balanced. We then transfer the concept of inhibitory and excitatory clustering of the connectivity to more realistic models of integrate and fire neurons and show that here it conserves the local balance of excitation and inhibition during cluster cycling dynamics.
Thereby the interval variability seen in cortical data is conserved during high activity states of the clusters. Finally we analysed a data set where monkeys executed a delayed centre out reach task where varying information about the target direction was given. The depth of modulation of the count variance in the monkey’s motor cortex was dependent on how much information was given during the delay period. Using a balanced network with inhibitory and excitatory clusters, we showed that the context dependent modulation of variability during the delay period can be captured by a simple probabilistic model of movement preparation.

Additional Information

PhD defence in the GRK 1589 "Sensory Computation in Neural Systems"

Organized by

Martin Nawrot / Robert Martin