PySpike
PySpike copied to clipboard
mariomulansky
Spike Sync matrix not symmteric
I see that the Spike Sync matrix is unsymmetric sometimes. What is this supposed to mean?
The matrix should always be symmetric, can you provide an example where it isn't?
Ok I figured it out. In one of my conditions since there are no spikes the matrix has 'nan' values.
Ah OK, I see. However, since the latest version, PySpike should also be able to handle empty spike trains without creating NaNs. See e.g.: a35402c208bd0ad31e5e60b6ddc55a3470e7bdde Do you work with PySpike 0.3? If yes and you still get NaN please provide me an example.
Hmm interesting. I just downloaded PySpike this week. So there is definitely something going on. Will try to look into this.
Yes I am using 0.3. For now please let me know if I need to set the NaNs to 0 or 1 according to the paper. I will try to get an example later.
Is the interpretation valid? Is it right to say that two neurons not firing are synchronized?
It is definitely arguable. After some discussion we arrived at this interpretation, with the main argument being that also the absence of spikes contains information. However, if this interferes with your analysis we can try to make another interpretation available.
I might have found the problem. It appears when providing some interval to the spike_sync_matrix function. I've committed a fix to develop 0d7255c
That is great! Also I just want to confirm that the max_tau is in seconds correct? So a max_tau of 5 ms should be written as max_tau=0.005
PySpike doesnt use dimensions. So max_tau should be given in the same dimension as the times in your spike trains.
To confirm: Did you try if the fix works for you? Then I would close this issue.
I haven't. I installed using pip install and so cant do a git pull and try it out at the moment. I have a deadline to meet soon. Will update once it is done.
Nope I still get NaNs. Its not a big deal for me I just set it to 1.
Hrm ok, are you using the latest version from Github? Can you provide an example so I can check into this?
We checked this with the latest version of PySpike and this issue has been resolved. This includes the case of empty spikes or spike trains with just one spike (see also the closed issue #29). We also double-checked that it works when providing some interval to the spike_sync_matrix function.
