Reader Comments

Post a new comment on this article

a model for latency tuning curves

Posted by tmasquelier on 19 Jun 2012 at 16:53 GMT

Congrats! This is solid evidence for latency-based coding of orientations in V1, which was scarce so far (mainly [32]).

I especially like the scheme in which the stimulus orientation is given by the preferred orientation of the first cell (or group of cells) to fire one (or n) spike(s) (Fig 7).
Contrast invariance comes for free: changing the contrast will shift the latencies (as observed experimentally in [26] but also in Albrecht, 2002 http://jn.physiology.org/...), but the relative latencies (or ranks) will be largely unaffected.

You maybe interested in the model I presented in http://dx.doi.org/10.1007...
Essentially I showed that:
1) Latency tuning curves (as in your Fig 3C) emerge naturally in a Hubel&Wiesel-like feedforward model whose connectivity is shaped by STDP.
2) Relative latencies are more reliable than absolute ones, especially when contrast is varied, or when you consider saccades (as opposed to simulus onsets).
Your data is fully consistent with this model :-)

Best,

Timothée Masquelier

No competing interests declared.