brendonw1
commented
6 years ago This is difficult and certainly many of us have handled this in handling our own datasets. But before we walk about outputs, can you explain how you're handling the input side? Cellarray of basepaths? And then it loops and calls itself using a single basepath input?
As for output - two thought
1) it may not be totally necessary to make single array for people - instead you can just use the multi-input to make sure it executes in the usual way but just on each out put... so like a batch processing but not necessarily to gather them all together. In that case you can write a secondary "Gather" function. A system like this is actually what I'd assume the multi-basepath input does... just batching without combining
2) If you want to combine, I like the principle of combining to the max extent that logically makes sense. So the most combined is to make a numeric array... but maybe if they have different dimensions or somthing you do a cell/struct. Then I'd make a metadata thing labeling, say, how many rows are from each session.
My thoughts Thanks for doing this
B
On Fri, May 18, 2018 at 1:51 PM, Dan Levenstein notifications@github.com wrote:
I'm adding functionality to bz_LoadCellinfo (which can then be copied to the other I/O functions) to load from an entire dataset of basePaths
Why? I calculated ISI statistics (function for this coming soon) for cells from a bunch of recordings, and saved as baseName.ISIStats.cellinfo.mat in each basePath... I would now like to load those from ALL basePaths in the dataset.
My question for y'all is: how would we like this to be output.
- Should it just return a 1/N structure array https://www.mathworks.com/help/matlab/matlab_prog/create-a-structure-array.html where: ISIStats(1) is the cellinfo structure from the first recording ISIStats(2) is the cellinfo structure from the second recording ... ISIStats(N) is the cellinfo structure from the Nth recording. These are a little unwieldy, and will be need to be combined post-hoc for easy comparison.
Or 2) should it be a single structure where each of the fields are concatenated? This is a little harder to code up and will inevitably have some bugs to iron out relating to alignment etc. But will be much easier to analyze.
or 3) other
Either way, it will tack on the baseName from which the cellInfo was loaded, so that things don't get lost. I.e. each cell will have a UID (unitID) and baseName, both of which will be needed to uniquely identify that cell. We can make a function to align two multi-basePath structures so that UIDs and baseNames line up.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/buzsakilab/buzcode/issues/198, or mute the thread https://github.com/notifications/unsubscribe-auth/ADXrTeNqbm96SxQb60FZDPoS0WSvplpLks5tzwoKgaJpZM4UFHm9 .
dlevenstein
petersenpeter
I'm adding functionality to bz_LoadCellinfo (which can then be copied to the other I/O functions) to load from an entire dataset of basePaths
Why? I calculated ISI statistics (function for this coming soon) for cells from a bunch of recordings, and saved as baseName.ISIStats.cellinfo.mat in each basePath... I would now like to load those from ALL basePaths in the dataset.
My question for y'all is: how would we like this to be output.
1) Should it just return a 1/N structure array where: ISIStats(1) is the cellinfo structure from the first recording ISIStats(2) is the cellinfo structure from the second recording ... ISIStats(N) is the cellinfo structure from the Nth recording. These are a little unwieldy, and will be need to be combined post-hoc for easy comparison.
Or 2) should it be a single structure where each of the fields are concatenated? This is a little harder to code up and will inevitably have some bugs to iron out relating to alignment etc. But will be much easier to analyze.
or 3) other
Either way, it will tack on the baseName from which the cellInfo was loaded, so that things don't get lost. I.e. each cell will have a UID (unitID) and baseName, both of which will be needed to uniquely identify that cell. We can make a function to align two multi-basePath structures so that UIDs and baseNames line up.