Tim Gollisch: Efficient coding versus feature detection in the retina
| When |
May 31, 2023
from 12:15 PM to 01:00 PM |
|---|---|
| Where | Bernstein Center, Hansastr. 9a, Lecture Hall. Hybrid Format via Zoom. Meeting ID and password will be sent with e-mail invitation. You can also contact Fiona Siegfried for Meeting ID and password. |
| Contact Name | Fiona Siegfried |
| Add event to calendar |
 vCal iCal
|
Abstract
The efficient coding hypothesis is a fundamental conceptual idea for how neural systems encode sensory information. As original proposed by Barlow in the 1960s, it posits that neuronal populations use as little spiking activity as possible for transmitting their sensory messages. For the vertebrate retina, this hypothesis has famously been applied to relate the center-surround structure of receptive fields to the task of reducing the redundancy inherent in natural images, thereby decorrelating the activity among retinal ganglion cells. Yet, only few experimental studies have probed efficient coding and decorrelation in the retina with natural stimuli, especially when considering not only natural image structure, but also natural temporal dynamics.
In this talk, I will discuss recordings of spiking activity from retinal ganglion cells of marmosets and mice in response to images that were shifted across the retina according to species-specific gaze trajectories. This revealed that many types of ganglion cells display only little decorrelation and, instead, transiently violate classical efficient coding by joint activity patterns. I will further present methods for investigating these activity patterns with computational models of nonlinear receptive fields. Finally, I will use the example of direction-selective retinal ganglion cells to connect the nonlinear processing and the resulting correlated responses to the task of robustly encoding specific stimulus features.