2:00pm, Tuesday, 14 Aug 2007:

<center><b>
A theory of robust population coding in the retina
</b>

<p>
Eizaburo Doi
<br>

</center>

It has long been suggested that one important goal of the retina is to
transfer the maximum amount of visual information to higher visual
areas. However, it remains unclear whether redundancies in the code
should merely be reduced for coding efficiency, or also be utilized
for coding reliability. Here we propose a framework of the optimal
linear coding that simultaneously counteracts image distortions and
neural noise using an arbitrary number of neurons, and show how
redundancies could be useful for the subsequent stages to decode the
original signal. We provide theoretical predictions of population
coding at different human retinal eccentricities, but the model can be
applied to the general problem as to how a population of neurons
cooperates to robustly represent sensory information.