Reader Comments
Post a new comment on this article
Post Your Discussion Comment
Please follow our guidelines for comments and review our competing interests policy. Comments that do not conform to our guidelines will be promptly removed and the user account disabled. The following must be avoided:
- Remarks that could be interpreted as allegations of misconduct
- Unsupported assertions or statements
- Inflammatory or insulting language
Thank You!
Thank you for taking the time to flag this posting; we review flagged postings on a regular basis.
closeHashing Collisions
Posted by derekjames on 07 Jun 2008 at 16:04 GMT
The paper states:
"We also use local inhibition between different prototype cells: when a cell fires at a given position and scale, it prevents all other cells from firing later at the same scale and within an s/2 × s/2 square neighborhood of the firing position. This competition, only used in the learning phase, prevents all the cells from learning the same pattern. Instead, the cell population self-organizes, each cell trying to learn a distinct pattern so as to cover the whole variability of the inputs."
As I understand it, you're implementing a method for making sure that hashing collisions do not occur, i.e. that cells are uniquely recruited for sensitivity to distinct patterns. You say that when a cell fires, it prevents all other neighboring cells from firing "later." Does this mean at all times in the future? If so, how might such a mechanism be implemented in real nervous systems? Is such a mechanism plausible?
RE: Hashing Collisions
tmasquelier replied to derekjames on 09 Jun 2008 at 15:45 GMT
Strictly speaking we prevent more than one cell firing in a given spatial neighborhood. This does not guarantee they learn different patterns, because the same pattern could appear at multiple location in the image. But in practice, it is often enough to ensure that the cells do indeed learn different patterns.
"Later" means for the current spike volley (recall we use an image-by-image propagation scheme). When the next image is propagated, all the cells are again allowed to fire. As said in the discussion, spike volleys may correspond to microsaccades and/or to oscillations, so "later" means for the next few tens of ms - something that inhibition could do.
You may be interested in our recent PLoS ONE paper on STDP in continous mode, as opposed to discrete spike volleys like here:
Masquelier T, Guyonneau R, Thorpe SJ (2008). Spike Timing Dependent Plasticity Finds the Start of Repeating Patterns in Continuous Spike Trains. PLoS ONE 3(1): e1377 doi:10.1371/journal.pone.0001377
I hope this helps,
Timothee Masquelier