Edinburgh Research Explorer

Anticipation in a population-coding system: a model of the inputs to rodent head direction cells

Research output: Contribution to conferencePoster

Original languageEnglish
Publication statusPublished - 2006
EventAREADNE 2006 Research in Encoding and Decoding of Neural Ensembles - Nomikos Conference Centre, Santorini, Greece
Duration: 22 Jun 200625 Jun 2006

Conference

ConferenceAREADNE 2006 Research in Encoding and Decoding of Neural Ensembles
CountryGreece
CitySantorini
Period22/06/0625/06/06

Abstract

The rodent head-direction (HD) system can be seen as one of several brain systems across different modalities and species that display anticipatory or predictive properties. Like the HD system, several of these systems process a continuous quantity like speed, direction or position using population codes: examples include motion perception extrapolation, motor control of hand movements, and hippocampal place cells. The HD system represents a one-dimensional variable (direction) so is relatively simple, and as such makes a good model for studying sensory integration and anticipation in a population-coding system.
Here, we simulate possible mechanisms underlying its anticipatory properties, using a large (~17 hours) corpus of rat tracking data for testing and validation. Previous explanations of HD anticipation have relied on putative offset connections between attractors, for which experimental evidence is so far lacking, and the developmental mechanisms needed to support such connectivity are unclear. We provide an alternative hypothesis based on the properties of the inputs to the HD system: we show that spike rate adaptation (SRA) and post-inhibitory rebound firing (PRF), as found in medial vestibular nucleus neurons in vitro, can generate realistic ATIs in a simulated HD attractor network. In the single attractor case, anticipation is always accompanied by an increase in tracking error, although it is small compared to simply increasing the gain of the input. However, when the HD signal is propagated trough several layers, optimal tracking in the final layer is achieved by generating anticipation in the first layer. We suggest experiments to test this idea.
In addition to the main result that input properties could be the source of HD anticipation, there are interesting observations to be made on the anticipatory time intervals (ATIs) of individual simulated cells. ATIs of different cells in the same brain area are known to be highly variable; we show that this appears to require little, if any, biological differences between cells, independent of the method used to generate them. A further source of ATI variability is the HD profile being tracked: different profiles lead to different ATIs for different anticipation models.

Event

AREADNE 2006 Research in Encoding and Decoding of Neural Ensembles

22/06/0625/06/06

Santorini, Greece

Event: Conference

ID: 18651480