[Comp-neuro] High level abstractions in concept cells;
Single spiking neurons have "meaning" and are actually at the
geltdavid at yahoo.com
Mon Aug 22 05:18:16 CEST 2011
Ethics aside, here are a few thoughts on the subject:
>When you record 10^3 neurons, have complex random temporal patterns and use
>statistics with an ambiguous measure (firing rate) that *provides little >information* everything becomes possible
Prove that the firing rate provides little information.Start with the relation between STDP and firing rate.
>everything becomes possible
>the firing rate measure *is not sensitive enough to detect these subtle aspectsPerhaps some other mechanism is used beside the rate.
>Very important, a strong firing rate to JA may indicate a reorganization >triggered when JA is repeatedly presented not JA concept
So where does the spider fits in all of this? Did the spider bit Jennifer Aniston during reoganization, or did Jennifer Aniston took a bite of the spider
before reorganization? So firing rate reduction is positively correlated with biting. Now what do you mean by reduction: three or four bites or just two bites would be enough?
>“Cracking” the neural code was not the main goal goalOf course not. You were looking to crack crack the temporal code.
>Based on firing rate one cannot >distinguish between spider and JA (the firing rate is 8Hz )Unless different sets of synaptic weights are used for spider and for Anniston.
>Concept cells are comparable to place cells in rodents
If place cells ensembles describe the place what would concept cells ensembles describe?
>During learning >these "expert" neurons reduce their firing rate. After one week of training >the neurons generate only few spikes between the tone and turn starts.One week. It takes one week for such a neuron to get above the threshold. What happens in between that time? Where are the concepts stored?
>The outcome in spike directivity is a counterexample for temporal codingNo, it is not. It probably is another view of a mechanism sub-serving the same function.
>You only need one single >counterexample to throw down a "solid"theoretical construct of temporal >coding.You definitely need more than one punch. I am on the ground, I reckon... Eureka, I found my glasses...
>After learning all these cells provide an efficient response with only >few spikes for the same eventCan it be that synaptic weights are within proper parameters to establish communication? No, it can not be. That would have to be correlated with the firing rate coding and that's not possible. Temporal code just got beaten while going to its place.
>The concept cells were found in different regions. For example, >“James Brolin” in right hippocampus, “Venus Williams” in left hippocampus, >“Marilyn Monroe” in left parahippocampal cortex, “Michael Jackson” in right >amygdala.*Maybe they are all going to the zoo to see the hippos on campus.I thought that memory consolidationoccurs somewhere else, the cortexperhaps, cause to need one week "to learn" about someone is way too much.
Yawn. Going to sleep now, hopefully no consolidation will occur tonight.
University of South Dundas
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Comp-neuro