"zero-size array" error with circular plot about netchos HOT 6 CLOSED

Hi @khashayar854,

Can you show me the code please? You might need to transform it to a pandas dataframe. Something like this :

df = pd.DataFrame(coh, index=ch_names, columns=ch_names)
circular(df)

from netchos.

Yes I had. Circular used to work with my own connectivity function. But it seems that it doesn't like MNE connectivity for some reason?

th = np.percentile(am, 95)
# set to nan every non-significant links
am[am < th] = np.nan
df = pd.DataFrame(am, index=chans, columns=chans) # am is the adjecency matrix, chans is the list of channels

fig = circular(df)
fig.show()

from netchos.

Maybe netchos has a different convention in the orientation of the upper / lower triangle. Could you try the two following tests:

# number of significant links
print(np.sum(~np.isnan(am)))

# try to transpose the array
df = pd.DataFrame(am.T, index=chans, columns=chans)

from netchos.

Yes, thank you.
The orientation of the upper/lower triangle was indeed the issue, and transposing the adjacency matrix solved it.
Of course, heatmap() didn't have any issue with the orientation of the triangle since it could accept either, not circular() though.
maybe checking the orientation of the triangle can be added to the circular function so that it transposes df automatically if needed, so it would directly work with MNE with no extra step too?
Thanks for the tip.

from netchos.

Yes, indeed, netchos should perform this test. I didn't thought about this before. If I remember correctly, I think scipy has functions to check if symmetrical arrays are corrects. Could you check it?

from netchos.

Sorry, python beginner here. Not sure if there is a built-in function for it, but it seems like people have some suggestions here:
https://stackoverflow.com/questions/42908334/checking-if-a-matrix-is-symmetric-in-numpy

from netchos.