_LinAlgError: linalg.eigh: The algorithm failed to converge because the input matrix is ill-conditioned or has too many repeated eigenvalues (error code: 994).

[holydick99]

after run my dataset there are sth worng, and i can't find the solution on the internet. So sad

[minD9it]

I have a same issue, did you solve it??

[holydick99]

I have a same issue, did you solve it??

the problem is there are too many 0 or 1 in my arrary, after i get mean in different channel it can run

[minD9it]

I have a same issue, did you solve it??

the problem is there are too many 0 or 1 in my arrary, after i get mean in different channel it can run

Was there any problem with the function 'zscore'?? I think my problem is in zscore function,

def zscore(a, axis=0, ddof=0):
    mns = a.mean(dim=axis)
    sstd = a.std(dim=axis, unbiased=(ddof == 1))
    if axis and mns.dim() < a.dim():
        x=(a - mns.unsqueeze(axis)).div(sstd.unsqueeze(axis)) #  **<<after this code 1**
        return x.masked_fill(torch.isnan(x), 0) # nan -> 0
        #return torch.nan_to_num((a - mns.unsqueeze(axis)).div(sstd.unsqueeze(axis)))
    else:
        x=a.sub_(mns).div(sstd) #  **<<after this code 2**
        return x.masked_fill(torch.isnan(x), 0) # nan -> 0
        #return torch.nan_to_num(a.sub_(mns).div(sstd))

after these code, variable x contains 'inf' or '-inf'. It causes the error :(

I can't find out solution about it...

[holydick99]

It's seems without any ERROR. I only can give you my view, the problem you are facing is likely due to division by zero or you fill too many 0 in your data, you can try replace NAN by mean. I'm just year 2 student of university so maybe my solutions are not so correctly