go back

Information Processing with Spiking Neurons Under the Control of Gamma Oscillations

Tobias Rodemann, "Information Processing with Spiking Neurons Under the Control of Gamma Oscillations", 2003.


In this work the functional role of neocortical neurons interacting with gamma oscillations for information processing in the cortex is analysed. For this investigation a specific functional system is selected: the fast feedforward recognition in the ventral visual system. There exists extensive experimental data on connectivity, neural responses, and the time courses of activity in different submodules. Looking at this data from the point of view of the neural coding problem we discover that a new coding scheme is required to satisfy the constraints brought up by neural hardware and the fast responses observed in animals and humans. A new coding scheme is proposed a that satisfies these constraints: oscillation-relative latency-coding. This scheme transmits information in the timing of single spikes, which is both much faster and more efficient than the standard rate-coding scheme. The model employs (evoked) gamma oscillations to provide an iterated, rhythmic reference signal. These oscillations are omnipresent in the cortex but so far no generally accepted functional model has been brought up. To show that these oscillations enable a latency encoding and decoding a detailed study of the neural circuit that contains a spiking neuron coupled to an oscillatory signal is conducted. Employing both theoretical analysis and computer simulations it is investigated how the neuron maps input variables to a single output spike. This operation is highly complex but can be controlled by the oscillation signal. Interpreting the oscillation as a functional control element, information can be gated or blocked, the input-output mapping can be modified, and the latency-coding scheme can be activated or inactivated by just changing oscillation parameters. It turns out that going from a rate-coding to a single spike latency-coding model does not only change the carrier of information but also the processing operation within a node. The analysis suggests a different type of processing operation in neocortical neurons (at least for the fast feedforward recognition) than previous approaches and hints at a central role of oscillations for the control of information processing.

Download Bibtex file Per Mail Request