Bandit session concatenation

[Taaseen Rahman]

Hi,

I've often experienced some of my FEDs just automatically restarting on me (and I really haven't had the opportunity to take a deeper look into why this happens). Usually I just concatenate csv files when this happens and for most session types it works fine. However, for bandit sessions, it's a bit troublesome. 

The GUI I use to process my csv files defines a block based on the block pellet counter. If I'm extremely unlucky, sometimes the FED will restart in the most awkward of times (e.g. the high probability is right poke, but the block pellet is 0, and then it auto-restarts, and the high probability poke is the left poke, and the block pellet resets to 0). This throws off the GUI I use because it thinks there's two high probability pokes in one block, and that shouldn't be possible technically. Would anyone have any ideas on how I might circumvent this issue? I've unfortunately accepted that I'll run into some FED auto-restarts for now, and I'm just trying to figure out if I can adjust either my concatenation code, or my GUI, or somehow delete incomplete blocks manually or something. Any help would be appreciated.

Sincerely,

Taaseen

[Lex]

Hi Taaseen,

I think the best way to deal with this is to delete the incomplete blocks at the end of the files before concatenating.  That way you are only analyzing complete blocks.  

If you're concerned about having two blocks in a row with the same side high you can delete the first block after the reset as well. 

Essentially my logic on this is to only analyze high quality data, so if a FED resets and this interferes with the task logic or your analysis logic, slice out the bad data so you're only analyzing the good data.

--
You received this message because you are subscribed to the Google Groups "FEDforum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fedforum+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/fedforum/3c41fa45-7049-4a22-b01d-74b1f3db0e18n%40googlegroups.com.

[Taaseen Rahman]

Hi Lex,

Funnily enough I tried this previously thinking it’d do the trick but I think it messes up the time in the csv files somehow. I was wondering if that’s expected if I remove incomplete blocks and concatenate it afterwards. I will also take a look at the concatenation code if something’s wrong with the way it processes the time but I was wondering if anyone else may have also experienced something like this. 

Sincerely,

Taaseen

[Lex]

If you're opening the data and manually editing it in Excel and then resaving it could mess with the times. I avoid opening/resaving FED files in Excel, it often seems intent on corrupting dates and timestamps! 

But if you're importing the fed csv into pandas and filtering out rows from the data frame it will not mess with time values in the unfiltered rows.

I'd try to narrow down how it's messing with your dates and see if that gives you a clue to what's going wrong.  Best, -Lex

[Taaseen Rahman]

Oh I see! That actually makes sense. I’ll play around with the data and see if I can salvage some of my files through pandas. Thanks again for the suggestion, this will save me from heaps of trouble.

Sincerely,

Taaseen

[Matt Gaidica]

Taaseen, I'm curious how often you see it resetting? We've been considering merging a more stable fork that eliminates some memory leaks, but for most people, the leaks never accumulate faster than a battery lasts, so it's a self-solving issue.

[Taaseen Rahman]

Hi Matt,

I notice these auto-restarts happen mostly when I'm doing long term studies where I keep the FEDs plugged in (e.g. for at least over a week). It takes maybe about 10 days before I run into such problems, but sometimes I get really unlucky and I run into this problem within the first day or two of the actual experiment. It could be because of moisture build up in the room causing a short circuit or something too, but I haven't had an opportunity to test out whether it's due to memory leaks yet. 

--

You received this message because you are subscribed to the Google Groups "FEDforum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fedforum+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/fedforum/f2f5f944-106d-42ef-a6d8-7e15b082cbc1n%40googlegroups.com.