Transmission and transformation of spiking activity in neuronal networks
CBN (Computational Biology and Neurocomputing) seminars
Friday 28 November 2014
to 15:00 at
Arvind Kumar (CB/CSC/KTH)
Both structural and functional evidence suggest that brain is highly modular and organised as a network of networks. To exploit the modularity it is important that the activity can be propagated from one network to others in a selective manner. That is, information processing involves not only ‘transformation’ of the input by the recurrent network but also the ‘transmission’ of the output. This division of transmission and transformation is directly associated with the balance of feedforward and feedback interactions.
In the first part of my talk I will describe how properties of the feedforward input, input projection and the dynamics in the receiving network affect the transmission of spiking activity and argue for ‘computing in transit’. In the second part, I will discuss how incoming feedforward input interacts with the ongoing activity to shape the statistics of the evoked activity. I will provide support that feedforward input properties can explain the dynamic state/task dependent modulation of noise correlation and trial-by-trial variability.
Finally, (if time permits) I will give a preview of ongoing work in my group on control of the activity in spiking neuronal networks and models of brain diseases that are manifested at the level of altered brain dynamics (epilepsy, fear-conditioning).