Splitbit value assignment

General issues of interest both for network and
individual cell parallelization.

Moderator: hines

Krishna Chaitanya
Posts: 70
Joined: Wed Jan 18, 2012 12:25 am
Location: University of Pavia

Splitbit value assignment

Hi,

In https://senselab.med.yale.edu/modeldb/S ... \split.hoc, what does splitbit refer to? How is it calculated? I didn't understand on what basis they are assigning splitbit values.

Can anyone kindly let me know. Thank you.

mctavish
Posts: 74
Joined: Tue Mar 14, 2006 6:10 pm
Location: New Haven, CT

Re: Splitbit value assignment

Thanks for the interest in our model.

Basically, our mitral cell is split into three parts: 1) the central Tuft, primary dendrite, soma and axon, 2) left dendrite, 3) right dendrite. Each part must have a unique GID. This is our hashing function to map these sections to GIDs.

We could have made a scheme to assign GIDs such that Mitral cell #1 = (0,1,2), Mitral cell #2 = (3,4,5), etc. Instead, we made the central part of each mitral cell have a "base GID" = 0,1,2, ..., #num mitrals-1. Our granule cells then start at GID=#num mitral and go to #nummitral + #numgran - 1. Now, assuming that this net will have fewer than splitbit=6e8 cells, we can assign the left dendrite = baseGID + splitbit and the right dendrite baseGID + 2*splitbit. This scheme lets us know at a glance that depending on the value of the number if it is a central segment, a left dendrite, or a right dendrite. Furthermore, given a left or right GID, doing a modulo on splitbit will retrieve the base GID. Perhaps most importantly, if the cell is not multisplit, then it keeps the whole cell assigned to the baseGID.

The particular value of 6e8 was chosen arbitrarily. We wanted a "large" number, but any number above nummitral+numgranule would suffice.

Krishna Chaitanya
Posts: 70
Joined: Wed Jan 18, 2012 12:25 am
Location: University of Pavia

Re: Splitbit value assignment

Dear mctavish,

Thank you very much for the in detail explanation about splitbit assignment in your model.