External Email - Use Caution
Hello
I was thing to do this example with my data, but in the end, I only get this error:
File "C:\Anaconda3\envs\mne\lib\site-packages\numpy\lib\function_base.py", line 2325, in cov
c = dot(X, X_T.conj())
MemoryError
What can cause it?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20190206/71c6c057/attachment.html
External Email - Use Caution
hi,
I don't think your error is related to MNE.
I suspect you want to compute cov(X.T) and not cov(X) ie
you swapped rows and columns
HTH
Alex
External Email - Use Caution
Well, I didn't change code much, only add my data.
I'm really confused where I made a mistake.
Now it looks this way:
data_path = 'C:/Users/Public/MEG/asami_ryo/170809/Asami1_raw.fif'
#print(raw_data)
#print(raw_data.ch_names[:204])
tmin, tmax = 0, 10.26
event_id = dict(l=1, r=2, n=4)
raw_data = mne.io.read_raw_fif(data_path, preload=True)
raw_data.filter(1.0, None, fir_design='firwin')
#raw_data.crop(tmin, tmax).load_data()
events = mne.find_events (raw_data)
picks = mne.pick_types(raw_data.info, meg=True, eeg=True, stim=False, eog=False)
epochs = mne.Epochs(raw_data, events, event_id=event_id, tmin=tmin, tmax=tmax, proj=False,baseline=(None, 0), preload=True,verbose=False)
labels = epochs.events[:, -1]
# get MEG and EEG data
meg_epochs = epochs.copy().pick_types(meg=True, eeg=False)
meg_data = meg_epochs.get_data().reshape(len(labels), -1)
clf = LogisticRegression(solver='lbfgs')
scaler = StandardScaler()
# create a linear model with LogisticRegression
model = LinearModel(clf)
# fit the classifier on MEG data
X = scaler.fit_transform(meg_data)
model.fit(X, labels)
# Extract and plot spatial filters and spatial patterns
for name, coef in (('patterns', model.patterns_), ('filters', model.filters_)):
# We fitted the linear model onto Z-scored data. To make the filters
# interpretable, we must reverse this normalization step
coef = scaler.inverse_transform([coef])[0]
# The data was vectorized to fit a single model across all time points and
# all channels. We thus reshape it:
coef = coef.reshape(len(meg_epochs.ch_names), -1)
# Plot
evoked = EvokedArray(coef, meg_epochs.info, tmin=epochs.tmin)
evoked.plot_topomap(title='MEG %s' % name, time_unit='s')
External Email - Use Caution
what is the size of meg_data for you? how does it compare with what is
used in mne example?
Alex
External Email - Use Caution
??? ??? ??? ??? ??? ??? ? ??? OneDrive. ??? ??? ???, ??? ??? ???.
<https://1drv.ms/u/s!AsTbwfRXq04Aa0PWbR2gdb0UdLU>
[https://r1.res.office365.com/owa/prem/images/dc-generic_20.png]<https://1drv.ms/u/s!AsTbwfRXq04Aa0PWbR2gdb0UdLU>
Asami1_raw.fif<https://1drv.ms/u/s!AsTbwfRXq04Aa0PWbR2gdb0UdLU>
I didn' checked the sample data of this example, just add mine raw data. It's one session with one person, so I assumed I can't it worse.
External Email - Use Caution
please check first with sample data. If you have a pb with sample you have
a setup / hardware limitation.
Alex
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20190206/effbe745/attachment-0001.html
External Email - Use Caution
Tried, but it works well with a sample data.
So it's be because I'm using a to big raw file? I need crop it? Or it's because of different filter's settings?
External Email - Use Caution
I cannot guess.
I need to know the size of the new data.
Alex
External Email - Use Caution
I think it s because you reshape your meg data (n_trials, n_sensors,
n_times) into a matrix (n_trials, n_sensors * n_times)
meg_data = meg_epochs.get_data().reshape(len(labels), -1)
The covariance of this matrix will be pretty heavy.
JR
External Email - Use Caution
It's 278 mb heavy, around 6 minutes.
External Email - Use Caution
I did reshape because it's in example, if it's too heavy I should refuse this conversion?
External Email - Use Caution
what is n_times for you?
basically how many samples and features do you have?
A
External Email - Use Caution
n_times is 348000 and total projection items are 10.
Probably there's a very simple solution, but I'm quite new for all this.
External Email - Use Caution
learning in dimension 348000 with 10 samples is not realistic.
You need more sample and to reduce the feature space / dimension.
Alex
