Reader Comments

Post a new comment on this article

Time component - what is important i sthe gradient of activity

Posted by Yehouda on 03 Jun 2008 at 19:54 GMT

I think the model misses the time component. What is recognized is not activity above some threshold, but a fast change in activity over short period (few tenths of a second). In other words, it is the _gradient_ of activity that have to be above some threshold, not the activity itself.

To recognize gradient the level of activity itself need to be routed through two (or more) routes, the fast activating and the slower inhibiting.

the advantage of this is that you don't need a threshold which depends on the level of activity, so recognition of a match can happen independently of the level activity in the system.

Yehouda Harpaz
129 Corrie Road
Cambridge CB1 3QQ

RE: Time component - what is important i sthe gradient of activity

laoptican replied to Yehouda on 13 Aug 2008 at 14:43 GMT

Our study identifies one of the sites of, and a synaptic mechanism for, short-term visual memory. We have shown that a neuronal signature consistent with a multiplicative memory mechanism exists in area TE of IT cortex. Our data do not support any particular mechanism by which downstream areas could decode the information contained in the spike trains. Our model does not include temporal modulation, as it is designed to support the multiplicative hypothesis of synaptic memory. A more elaborate model, with temporal dynamics, would be needed to test ideas about decoding.