Tick-Tock: How Global Oscillatory “Clocks” May Support Computation In The Brain

David Markowitz, Princeton University

Photo of David Markowitz

In many regions of the brain, neurons represent analog values by firing action potentials at distinct mean rates. This “rate coding” scheme underlies most models of biological neural networks and the whole of artificial neural network theory. A problem arises, however, when a neuron is tasked with comparing analog values represented by two or more of its input neurons. For example, using standard modeling techniques it is nontrivial for a cell to detect when many analog input values are equal, such as may be required for concentration invariant odorant detection in the olfactory system.

Previous theoretical work has demonstrated that a common sinusoidal drive given to all neurons in a population leads to near-perfect synchronization of cells firing at similar rates. Detection of synchronous activity (which is straightforward using conventional neuron models) then corresponds to a Many-Are-Equal (MAE) computation, because only cells firing at similar rates will be synchronized. In this manner, a global clock forces neurons to convey information of relevance to MAE computation in the temporal domain while preserving information about analog values in mean firing rates. Thus, a global clock facilitates multi-cell analog computation.

Consistent with the above, real brains are known to exhibit spatially coordinated oscillatory dynamics during active sensation, and individual neurons have been observed to receive synaptic input that is synchronous with oscillatory population activity. Contrary to previous theoretical work, however, this rhythmic drive is not sinusoidal, but instead contains power distributed across a broad range of frequencies. This raises the question of whether MAE computation remains feasible when global background oscillations are noisy.

In the present study, I explore this issue and establish constraints on background oscillations for MAE computation to be feasible in a population of neurons.

Abstract Author(s): David A. Markowitz, David W. Tank, Department of Molecular Biology, Princeton University