vazirilab / sid Goto Github PK

@jmmanley wirtes:
SID_output.neur_rad: Might be worth overall checking on this parameter, as it is set multiple times in either Input, SID_output, and opts. There is e.g. a hard-coded value opts.neur_rad = 6.

In the paper, the reconstruction is performed on the std image. However, in the code, the reconstruction is performed on the spatial filter computed from NNMF, and I could not find the place where the std image is used. Could you explain this?

An error occur at the "%% Extract time series at location indir" section of the sid_main where the "incremental_temporal_update_gpu" is called. In my case the sparse array size of [12837x667785 sparse double] (down from 69GBytes to 1GBytes sparse vs. full array versions) could not be handled.
Just deactivated the "gpu_ids" for this function ONLY.
A more permanent solution would probably be to approach the array split-wise.