def ml_classifier(inputs, targets, classifier=None, pipeline=None):
"""Uses sklearn to fit a model given inputs and targets
Args:
inputs: list containing (N trials * M channels) data segments of length(number of features).
targets: list containing (N trials * M channels) of marker data (0 or 1).
classifier: pre-trained lda classifier; if None train from scratch
pipeline: name of pipeline to create if classifier is None
Returns:
classifier: classifier object
"""
pipeline_dict = {
'vect_lr':
make_pipeline(Vectorizer(), StandardScaler(), LogisticRegression()),
'vecct_reglda':
make_pipeline(Vectorizer(), LDA(shrinkage='auto', solver='eigen')),
'xdawn_reglda':
make_pipeline(Xdawn(2, classes=[1]), Vectorizer(),
LDA(shrinkage='auto', solver='eigen')),
'erpcov_ts':
make_pipeline(ERPCovariances(), TangentSpace(), LogisticRegression()),
'erpcov_mdm':
make_pipeline(ERPCovariances(), MDM())
}
if not classifier and pipeline:
classifier = pipeline_dict[pipeline.lower()]
classifier.fit(inputs, targets)
return classifier
def N170_test(session_data):
markers = N170_MARKERS
epochs = get_session_erp_epochs(session_data, markers)
conditions = OrderedDict()
for i in range(len(markers)):
conditions[markers[i]] = [i+1]
clfs = OrderedDict()
clfs['Vect + LR'] = make_pipeline(Vectorizer(), StandardScaler(), LogisticRegression())
clfs['Vect + RegLDA'] = make_pipeline(Vectorizer(), LDA(shrinkage='auto', solver='eigen'))
clfs['ERPCov + TS'] = make_pipeline(ERPCovariances(estimator='oas'), TangentSpace(), LogisticRegression())
clfs['ERPCov + MDM'] = make_pipeline(ERPCovariances(estimator='oas'), MDM())
clfs['XdawnCov + TS'] = make_pipeline(XdawnCovariances(estimator='oas'), TangentSpace(), LogisticRegression())
clfs['XdawnCov + MDM'] = make_pipeline(XdawnCovariances(estimator='oas'), MDM())
methods_list = ['Vect + LR','Vect + RegLDA','ERPCov + TS','ERPCov + MDM','XdawnCov + TS','XdawnCov + MDM']
# format data
epochs.pick_types(eeg=True)
X = epochs.get_data() * 1e6
times = epochs.times
y = epochs.events[:, -1]
# define cross validation
cv = StratifiedShuffleSplit(n_splits=20, test_size=0.25,
random_state=42)
# run cross validation for each pipeline
auc = []
methods = []
print('Calcul in progress...')
for m in clfs:
try:
res = cross_val_score(clfs[m], X, y==2, scoring='roc_auc',
cv=cv, n_jobs=-1)
auc.extend(res)
methods.extend([m]*len(res))
except Exception:
print("exception")
## Plot Decoding Results
results = pd.DataFrame(data=auc, columns=['AUC'])
results['Method'] = methods
n_row,n_column = results.shape
auc_means = []
for method in methods_list:
auc = []
for i in range(n_row):
if results.loc[i,'Method']== method:
auc.append(results.loc[i,'AUC'])
auc_means.append(np.mean(auc))
counter = 0
for i in range(len(methods_list)):
color = 'green' if auc_means[i]>=0.7 else 'red'
counter = counter +1 if auc_means[i]>=0.7 else counter
return counter > 0, counter
def test_xdawn_decoding_performance():
"""Test decoding performance and extracted pattern on synthetic data."""
from sklearn.model_selection import KFold
from sklearn.pipeline import make_pipeline
from sklearn.linear_model import LogisticRegression
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import accuracy_score
n_xdawn_comps = 3
expected_accuracy = 0.98
epochs, mixing_mat = _simulate_erplike_mixed_data(n_epochs=100)
y = epochs.events[:, 2]
# results of Xdawn and _XdawnTransformer should match
xdawn_pipe = make_pipeline(
Xdawn(n_components=n_xdawn_comps),
Vectorizer(),
MinMaxScaler(),
LogisticRegression(solver='liblinear'))
xdawn_trans_pipe = make_pipeline(
_XdawnTransformer(n_components=n_xdawn_comps),
Vectorizer(),
MinMaxScaler(),
LogisticRegression(solver='liblinear'))
cv = KFold(n_splits=3, shuffle=False)
for pipe, X in (
(xdawn_pipe, epochs),
(xdawn_trans_pipe, epochs.get_data())):
predictions = np.empty_like(y, dtype=float)
for train, test in cv.split(X, y):
pipe.fit(X[train], y[train])
predictions[test] = pipe.predict(X[test])
cv_accuracy_xdawn = accuracy_score(y, predictions)
assert_allclose(cv_accuracy_xdawn, expected_accuracy, atol=0.01)
# for both event types, the first component should "match" the mixing
fitted_xdawn = pipe.steps[0][1]
if isinstance(fitted_xdawn, Xdawn):
relev_patterns = np.concatenate(
[comps[[0]] for comps in fitted_xdawn.patterns_.values()])
else:
relev_patterns = fitted_xdawn.patterns_[::n_xdawn_comps]
for i in range(len(relev_patterns)):
r, _ = stats.pearsonr(relev_patterns[i, :], mixing_mat[0, :])
assert np.abs(r) > 0.99
def svm_proba(train_xx6, train_y, test_xx6):
selects = []
for j, y in enumerate(train_y):
if y == 1:
[selects.append(j - e) for e in [-2, -1, 0, 1, 2]]
train_xx6 = train_xx6[selects]
train_xx6 = train_xx6[:, :, 40:80]
test_xx6 = test_xx6[:, :, 40:80]
train_y = train_y[selects]
train_y[train_y != 1] = 0
# s = train_xx6.shape
# train_xx6 = train_xx6.reshape((s[0], s[1] * s[2]))
# s = test_xx6.shape
# test_xx6 = test_xx6.reshape((s[0], s[1] * s[2]))
# return test_xx6
clf = make_pipeline(
Vectorizer(), StandardScaler(), pca.PCA(n_components=.95),
svm.SVC(gamma='scale',
kernel='rbf',
class_weight={
0: 1,
1: 2
},
probability=True))
clf.fit(train_xx6, train_y)
return clf.predict_proba(test_xx6)
def test_get_coef_multiclass_full(n_classes, n_channels, n_times):
"""Test a full example with pattern extraction."""
from sklearn.pipeline import make_pipeline
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import StratifiedKFold
data = np.zeros((10 * n_classes, n_channels, n_times))
# Make only the first channel informative
for ii in range(n_classes):
data[ii * 10:(ii + 1) * 10, 0] = ii
events = np.zeros((len(data), 3), int)
events[:, 0] = np.arange(len(events))
events[:, 2] = data[:, 0, 0]
info = create_info(n_channels, 1000., 'eeg')
epochs = EpochsArray(data, info, events, tmin=0)
clf = make_pipeline(
Scaler(epochs.info),
Vectorizer(),
LinearModel(LogisticRegression(random_state=0, multi_class='ovr')),
)
scorer = 'roc_auc_ovr_weighted'
time_gen = GeneralizingEstimator(clf, scorer, verbose=True)
X = epochs.get_data()
y = epochs.events[:, 2]
n_splits = 3
cv = StratifiedKFold(n_splits=n_splits)
scores = cross_val_multiscore(time_gen, X, y, cv=cv, verbose=True)
want = (n_splits, )
if n_times > 1:
want += (n_times, n_times)
assert scores.shape == want
assert_array_less(0.8, scores)
clf.fit(X, y)
patterns = get_coef(clf, 'patterns_', inverse_transform=True)
assert patterns.shape == (n_classes, n_channels, n_times)
assert_allclose(patterns[:, 1:], 0., atol=1e-7) # no other channels useful
def __init__(self, subject=None, date=None, mode='train', **kwargs):
if subject is None:
subject = cfg.subj_info.subjname
self._subj_path = os.path.dirname(__file__) + '/../data/' + subject
if date is None:
self._date = utils.find_nearest_time(self._subj_path)
else:
if isinstance(date, datetime):
# convert datetime to str
self._date = date.strftime("%Y-%m-%d-%H-%M-%S")
else:
self._date = date
self.mode = mode.lower()
assert self.mode in ['train', 'test']
if self.mode == 'test':
# loading trained coefficient
self.data_dict = np.load(os.path.join(self._subj_path, self._date, 'coef.npz'))
# loading trained model
self.__cls = joblib.load(os.path.join(self._subj_path, self._date, 'model.pkl'))
self._ch_ind = self.data_dict['ind_ch_scores']
else:
self.data_dict = {}
C = kwargs.pop('C', 1)
n_components = kwargs.pop('n_components', 3)
self.__cls = make_pipeline(
_XdawnTransformer(n_components=n_components),
ChannelScaler(),
Vectorizer(),
LogisticRegression(C=C, class_weight='balanced', solver='liblinear', multi_class='ovr')
)
def mvpa(name):
# Perform MVPA
# Setting
BASELINE = (None, 0)
CROP = (0, 0.8)
EVENTS = ['1', '2']
# Load epochs
loader = FileLoader(name)
loader.load_epochs(recompute=False)
print(loader.epochs_list)
# Prepare [predicts] for results
predicts = []
# Cross validation
num_epochs = len(loader.epochs_list)
for exclude in range(num_epochs):
# Start on separate training and testing dataset
print(f'---- {name}: {exclude} | {num_epochs} ----------------------')
includes = [
e for e in range(len(loader.epochs_list)) if not e == exclude
]
excludes = [exclude]
train_epochs, test_epochs = loader.leave_one_session_out(
includes, excludes)
print(train_epochs, test_epochs)
def prepare_epochs(epochs):
# A tool for prepare epochs
epochs = epochs['1', '2']
epochs.apply_baseline(BASELINE)
return epochs.crop(CROP[0], CROP[1])
# print('Xdawn --------------------------------')
# enhancer = Enhancer(train_epochs=train_epochs,
# test_epochs=test_epochs)
# train_epochs, test_epochs = enhancer.fit_apply()
# Prepare epochs
train_epochs = prepare_epochs(train_epochs)
test_epochs = prepare_epochs(test_epochs)
X_train, y_train = get_X_y(train_epochs)
X_test, y_test = get_X_y(test_epochs)
print('Training -----------------------------')
clf = svm.SVC(gamma='scale', kernel='rbf', class_weight='balanced')
pipeline = make_pipeline(Vectorizer(), StandardScaler(), clf)
pipeline.fit(X_train, y_train)
y_pred = pipeline.predict(X_test)
# y_pred = y_test
print('Testing ------------------------------')
predicts.append(dict(y_test=y_test, y_pred=y_pred))
with open(os.path.join(results_dir, f'{name}.json'), 'wb') as f:
pickle.dump(predicts, f)
pass
def UpdateModel():
try:
SetPathsVars()
userlist = np.load(USER_LIST, allow_pickle='TRUE').item()
if (len(userlist) < 2):
return True
database = np.load(DB_PATH, allow_pickle='TRUE')
print(database)
database = database[1:]
rows, cols = len(database), len(database[0])
print(rows, cols)
X = []
y = []
for r in database:
y.append(r[-1])
arrr = r[:cols - 1]
arrr = np.reshape(arrr, (15, 113))
X.append(arrr)
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.svm import LinearSVC
from mne.decoding import Vectorizer
clf = make_pipeline(Vectorizer(), StandardScaler(), LinearSVC())
clf.fit(X, y)
pkl.dump(clf, open(getPath('Res/model.pkl'), 'wb'))
print('Successfully updated model')
return True
except Exception as e:
print(e)
print("Update Model haga")
return False
def train_test_lda(self, X_train, y_train, X_test, y_test):
"""Regulated LDA
Parameters
----------
X_train: instance of numpy.ndarray
The training data.
y_train: instance of numpy.ndarray
The training target values.
X_test: instance of numpy.ndarray
The testing data.
y_test: instance of numpy.ndarray
The testing target values.
Returns
-------
model : instance of sklearn.pipeline.Pipeline
The final model.
auc : float
The AUC score.
"""
model = make_pipeline(Vectorizer(),
LDA(shrinkage='auto', solver='eigen'))
model.fit(X_train, y_train)
auc = roc_auc_score(y_test, model.predict(X_test))
return model, auc
def make_clf(pattern=False,vectorized=False):
clf = []
from sklearn.svm import SVC
clf.append(('vectorizer',Vectorizer()))
# use linear SVM as the estimator
estimator = SVC(max_iter=-1,kernel='linear',random_state=12345,class_weight='balanced',probability=True)
clf.append(('estimator',estimator))
clf = Pipeline(clf)
return clf
def mvpa(name):
# Perform MVPA
# Load epochs
loader = FileLoader(name)
loader.load_epochs(recompute=False)
print(loader.epochs_list)
# Prepare [predicts] for results
predicts = []
# Cross validation
num_epochs = len(loader.epochs_list)
for exclude in range(num_epochs):
# Start on separate training and testing dataset
print(f'---- {name}: {exclude} | {num_epochs} ----------------------')
includes = [
e for e in range(len(loader.epochs_list)) if not e == exclude
]
excludes = [exclude]
train_epochs, test_epochs = loader.leave_one_session_out(
includes, excludes)
print(train_epochs, test_epochs)
print('Xdawn --------------------------------')
enhancer = Enhancer(train_epochs=train_epochs, test_epochs=test_epochs)
train_epochs, test_epochs = enhancer.fit_apply()
# Prepare epochs
train_epochs = prepare_epochs(train_epochs)
test_epochs = prepare_epochs(test_epochs)
X_train, y_train = get_X_y(train_epochs)
X_test, y_test = get_X_y(test_epochs)
print('Preprocess ---------------------------')
pipeline = make_pipeline(Vectorizer(), StandardScaler())
X_train = pipeline.fit(X_train)
X_test = pipeline.fit(X_test)
y_train[not y_train == 1] = 0
y_test[not y_test == 1] = 0
print('Training -----------------------------')
eegnet = EEGNet_classifier()
eegnet.fit(X_train, y_train, quiet=False)
y_pred = eegnet.predict(X_test)
# y_pred = y_test
print('Testing ------------------------------')
predicts.append(dict(y_test=y_test, y_pred=y_pred))
with open(os.path.join(results_dir, f'{name}.json'), 'wb') as f:
pickle.dump(predicts, f)
pass
def make_clf():
clf = []
clf.append(('vectorizer',Vectorizer()))
# hyper parameters were optimized and here we just directly use them in the random forest model
clf.append(('estimator',RandomForestClassifier(n_estimators=190,# number of trees
max_depth=None, # no need to specifiy the depth, in order words, feature depth
random_state=12345,
class_weight='balanced',
max_features=10, # dimension reduction
min_samples_leaf=4,# minimum feature span
min_samples_split=4)))# minimum feature split
clf = Pipeline(clf)
return clf
def svm_fit_predict(train_xraw, train_y, test_xraw):
clf = svm.SVC(gamma='scale',
kernel='rbf',
class_weight='balanced',
probability=True)
selects = []
for j, y in enumerate(train_y):
if y == 1:
[selects.append(j - e) for e in [-1, 0, 1]]
pipeline = make_pipeline(Vectorizer(), clf)
pipeline.fit(train_xraw[selects, :, 40:80], train_y[selects])
pred = pipeline.predict(test_xraw[:, :, 40:80])
prob = pipeline.predict_proba(test_xraw[:, :, 40:80])
return pred, prob
def make_clf(pattern=False, vectorized=False):
clf = []
if vectorized:
clf.append(('vectorizer', Vectorizer()))
clf.append(('scaler', MinMaxScaler()))
# use linear SVM as the estimator
estimator = SVC(max_iter=-1,
kernel='linear',
random_state=12345,
class_weight='balanced',
probability=True)
if pattern:
estimator = LinearModel(estimator)
clf.append(('estimator', estimator))
clf = Pipeline(clf)
return clf
def test_vectorizer():
"""Test Vectorizer."""
data = np.random.rand(150, 18, 6)
vect = Vectorizer()
result = vect.fit_transform(data)
assert_equal(result.ndim, 2)
# check inverse_trasnform
orig_data = vect.inverse_transform(result)
assert_equal(orig_data.ndim, 3)
assert_array_equal(orig_data, data)
assert_array_equal(vect.inverse_transform(result[1:]), data[1:])
# check with different shape
assert_equal(vect.fit_transform(np.random.rand(150, 18, 6, 3)).shape,
(150, 324))
assert_equal(vect.fit_transform(data[1:]).shape, (149, 108))
# check if raised errors are working correctly
vect.fit(np.random.rand(105, 12, 3))
assert_raises(ValueError, vect.transform, np.random.rand(105, 12, 3, 1))
assert_raises(ValueError, vect.inverse_transform,
np.random.rand(102, 12, 12))
def test_get_coef_multiclass(n_features, n_targets):
"""Test get_coef on multiclass problems."""
# Check patterns with more than 1 regressor
from sklearn.linear_model import LinearRegression, Ridge
from sklearn.pipeline import make_pipeline
X, Y, A = _make_data(n_samples=30000,
n_features=n_features,
n_targets=n_targets)
lm = LinearModel(LinearRegression()).fit(X, Y)
assert_array_equal(lm.filters_.shape, lm.patterns_.shape)
if n_targets == 1:
want_shape = (n_features, )
else:
want_shape = (n_targets, n_features)
assert_array_equal(lm.filters_.shape, want_shape)
if n_features > 1 and n_targets > 1:
assert_array_almost_equal(A, lm.patterns_.T, decimal=2)
lm = LinearModel(Ridge(alpha=0))
clf = make_pipeline(lm)
clf.fit(X, Y)
if n_features > 1 and n_targets > 1:
assert_allclose(A, lm.patterns_.T, atol=2e-2)
coef = get_coef(clf, 'patterns_', inverse_transform=True)
assert_allclose(lm.patterns_, coef, atol=1e-5)
# With epochs, scaler, and vectorizer (typical use case)
X_epo = X.reshape(X.shape + (1, ))
info = create_info(n_features, 1000., 'eeg')
lm = LinearModel(Ridge(alpha=1))
clf = make_pipeline(
Scaler(info, scalings=dict(eeg=1.)), # XXX adding this step breaks
Vectorizer(),
lm,
)
clf.fit(X_epo, Y)
if n_features > 1 and n_targets > 1:
assert_allclose(A, lm.patterns_.T, atol=2e-2)
coef = get_coef(clf, 'patterns_', inverse_transform=True)
lm_patterns_ = lm.patterns_[..., np.newaxis]
assert_allclose(lm_patterns_, coef, atol=1e-5)
# Check can pass fitting parameters
lm.fit(X, Y, sample_weight=np.ones(len(Y)))
def test_vectorizer():
"""Test Vectorizer."""
data = np.random.rand(150, 18, 6)
vect = Vectorizer()
result = vect.fit_transform(data)
assert_equal(result.ndim, 2)
# check inverse_trasnform
orig_data = vect.inverse_transform(result)
assert_equal(orig_data.ndim, 3)
assert_array_equal(orig_data, data)
assert_array_equal(vect.inverse_transform(result[1:]), data[1:])
# check with different shape
assert_equal(vect.fit_transform(np.random.rand(150, 18, 6, 3)).shape,
(150, 324))
assert_equal(vect.fit_transform(data[1:]).shape, (149, 108))
# check if raised errors are working correctly
vect.fit(np.random.rand(105, 12, 3))
assert_raises(ValueError, vect.transform, np.random.rand(105, 12, 3, 1))
assert_raises(ValueError, vect.inverse_transform,
np.random.rand(102, 12, 12))
tmin = i * 1.0
tmax = (i + 1) * 1.0
#Create :class:'Epochs <mne.Epochs>' object
epochs = mne.Epochs(raw,
events=events,
event_id=event_id,
tmin=tmin,
tmax=tmax,
baseline=None,
verbose=True,
preload=True)
for i in range(0, len(epochs.events)):
if i % 2 == 0:
epochs.events[i, 2] = 3
#epochs.plot(scalings = 'auto',block = True,n_epochs=10)
X = epochs.pick_types(meg=False, eeg=True)
y = epochs.events[:, -1]
# Define a unique pipeline to sequentially:
clf = make_pipeline(
Vectorizer(), # 1) vectorize across time and channels
StandardScaler(), # 2) normalize features across trials
LinearModel(LogisticRegression())) # 3) fits a logistic regression
clf.fit(X, y)
coef = get_coef(clf, 'patterns_', inverse_transform=True)
evoked = EvokedArray(coef, epochs.info, tmin=epochs.tmin)
fig = evoked.plot_topomap(title='EEG Patterns', size=3, show=False)
fig.savefig(title + "_ti_" + str(tmin) + "_tf_" + str(tmax) + '.png')
return X, y
# %%
for idx in range(1, 11):
# Loading data ------------------------------------------
running_name = f'MEG_S{idx:02d}'
band_name = 'U07'
worker = MEG_Worker(running_name=running_name)
worker.pipeline(band_name=band_name)
# MVPA ----------------------------------------------------------------
# Prepare classifiers
_svm = svm.SVC(gamma='scale', kernel='rbf', class_weight='balanced')
clf = make_pipeline(Vectorizer(), StandardScaler(), _svm)
# Prepare paired X and y
# Set crop
crops = dict(a=(0.2, 0.4),
b=(0.4, 0.6),
c=(0.6, 0.8),
d=(0.2, 0.8),
e=(0.0, 1.0))
output_dict = dict()
for crop_key in crops:
crop = crops[crop_key]
# Get X and y for class 1
X1, y1 = pair_X_y(worker.clean_epochs, 1, crop)
# This approach classifies the data within, rather than across, subjects.
for chroma in ['hbo', 'hbr']:
st_scores = []
for sub in subjects:
bids_path = dataset.update(subject=sub)
raw_haemo, epochs = epoch_preprocessing(bids_path)
epochs.pick(chroma)
X = epochs.get_data()
y = epochs.events[:, 2]
clf = make_pipeline(Scaler(epochs.info), Vectorizer(),
LogisticRegression(solver='liblinear'))
scores = 100 * cross_val_multiscore(
clf, X, y, cv=5, n_jobs=1, scoring='roc_auc')
st_scores.append(np.mean(scores, axis=0))
print(f"Average spatio-temporal ROC-AUC performance ({chroma}) = "
f"{np.round(np.mean(st_scores))} % ({np.round(np.std(st_scores))})")
# %%
# Conclusion
# ----------
#
# Data were epoched then decoding was performed on the hbo signal and the hbr
def SVM_decoding_on_full_epochs(X,
y,
plot_conf_matrix=0,
class_names=None,
test_size=0.2,
n_splits=5):
""" This function decodes on the full epoch using standard SVM algorithm
Parameters
---------
X : data extracted from the epochs provided to the decoder
y : categorical variable (i.e. discrete but it can be more then 2 categories)
plot_confusion_matrix : set to 1 if you wanna see the confusion matrix
class_names: needed for the legend if confusion matrices are plotted ['cat1','cat2','cat3']
test_size : proportion of the data on which you want to test the decoder
n_splits : when calculating the score, number of cross-validation folds
Returns:
-------
score, y_test, y_pred
"""
# ------- define the classifier -------
scaler = preprocessing.StandardScaler()
vectorizer = Vectorizer()
clf = SVC(C=1, kernel='linear', decision_function_shape='ovr')
concat_classifier = Pipeline([('vector', vectorizer), ('scaler', scaler),
('svm', clf)])
# This returns the 5 scores calculated for each fold
kf = StratifiedKFold(n_splits=n_splits, shuffle=True, random_state=42)
y = np.asarray(y)
scores = []
for train_index, test_index in kf.split(X, y):
X_train, X_test = X[train_index], X[test_index]
y_train, y_test = y[train_index], y[test_index]
# Train on X_train, y_train
concat_classifier.fit(X_train, y_train)
# Test on X_test and then score
y_pred = concat_classifier.predict(X_test)
scores.append(accuracy_score(y_true=y_test, y_pred=y_pred))
scores = np.asarray(scores)
if plot_conf_matrix == 1:
print('you chose to plot the confusion matrix')
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=test_size, random_state=7, stratify=y)
y_pred = concat_classifier.fit(X_train, y_train).predict(X_test)
# Compute confusion matrix
cnf_matrix = confusion_matrix(y_test, y_pred)
np.set_printoptions(precision=3)
print(cnf_matrix)
# Plot non-normalized confusion matrix
plt.figure()
plot_confusion_matrix(cnf_matrix,
classes=class_names,
title='Confusion matrix, without normalization')
# Plot normalized confusion matrix
plt.figure()
plot_confusion_matrix(cnf_matrix,
classes=class_names,
normalize=True,
title='Normalized confusion matrix')
plt.show()
return scores, y_test, y_pred, cnf_matrix
return scores, cnf_matrix
def run_ica_experiment(_run, method_idx):
# filepaths = Path(r"C:\Users\paull\Documents\GIT\BCI_MsC\notebooks\BCI_Comp_IV_2a\BCICIV_2a_gdf/").glob("*T.gdf")
dataset, metadata = BCI_IV_Comp_Dataset.load_dataset(filepaths,
as_epochs=True,
concatenate=False,
drop_bad=True,
return_metadata=True,
tmin=-1.,
tmax=3.)
all_methods = get_all_methods()
methods = all_methods if method_idx is None else [all_methods[method_idx]]
name = "" if method_idx is None else "_{}".format(all_methods[method_idx])
print("Using methods", methods)
results = dict()
for method in methods:
print("Running for method", method)
clf = make_pipeline(
CSP(n_components=CSP_N_COMPONENTS), Vectorizer(), MinMaxScaler(),
LogisticRegression(penalty='l2', multi_class='auto'))
results[method] = list()
for i, (epochs, mdata) in enumerate(zip(dataset, metadata)):
print("\t", i, mdata["id"])
ICA = get_ica_instance(method, n_components=ICA_N_COMPONENTS)
start = time.time()
epochs = epochs.copy().load_data().filter(l_freq=None,
h_freq=40).resample(90.)
transformed_epochs = ICA.fit(epochs).get_sources(epochs)
duration = time.time() - start
scores = dict()
signal = np.hstack(transformed_epochs.get_data())
for fn_name in SCORING_FN_DICT:
score = apply_pairwise_parallel(signal,
SCORING_FN_DICT[fn_name])
scores[fn_name] = score
X, Y = transformed_epochs.get_data(), transformed_epochs.events[:,
2]
del epochs, transformed_epochs
try:
clf.fit(X, Y)
except Exception:
print("\t\tFailed during fit")
results[method].append({
"id": mdata["id"],
"score": None,
"bas": None,
"duration": duration
})
continue
pred = clf.predict(X)
bas = balanced_accuracy_score(Y, pred)
results[method].append({
"id": mdata["id"],
"score": scores,
"bas": bas,
"duration": duration
})
results_filepath = f"./results{name}.json"
with open(results_filepath, "w") as json_file:
json.dump(results, json_file, indent=4)
_run.add_artifact(results_filepath, content_type="json")
if subject_id in exclude:
continue
subject = 'S%02d' % subject_id
data_path = os.path.join('/home/claire/DATA/Data_Face_House/' + subject +
'/EEG/Evoked_Lowpass')
fname_in = os.path.join(data_path, '%s-epo.fif' % subject)
epochs = mne.read_epochs(fname_in)
epochs.interpolate_bads()
# all_epochs.append(epochs)
#epochs = mne.concatenate_epochs(all_epochs)
epochs.pick_types(eeg=True)
# Create classification pipeline
clf = make_pipeline(Xdawn(n_components=3, reg='oas'), Vectorizer(),
MinMaxScaler(), LogisticRegression(penalty='l1'))
le = LabelEncoder()
labels = le.fit_transform(epochs.events[:, 2])
# Cross validator
cv = StratifiedKFold(y=labels, n_folds=10, shuffle=True, random_state=42)
# Do cross-validation
preds = np.empty(len(labels))
for train, test in cv:
clf.fit(epochs[train], labels[train])
preds[test] = clf.predict(epochs[test])
# Classification report
exclude='bads')
epochs = Epochs(raw,
events,
event_id,
tmin,
tmax,
proj=False,
picks=picks,
baseline=None,
preload=True,
verbose=False)
# Create classification pipeline
clf = make_pipeline(
Xdawn(n_components=n_filter), Vectorizer(), MinMaxScaler(),
LogisticRegression(penalty='l1', solver='liblinear', multi_class='auto'))
# Get the labels
labels = epochs.events[:, -1]
# Cross validator
cv = StratifiedKFold(n_splits=10, shuffle=True, random_state=42)
# Do cross-validation
preds = np.empty(len(labels))
for train, test in cv.split(epochs, labels):
clf.fit(epochs[train], labels[train])
preds[test] = clf.predict(epochs[test])
# Classification report
tmax=0.8,
baseline=None,
reject={'eeg': 75e-6},
preload=True,
verbose=False,
picks=[0, 1, 2, 3])
print('sample drop %: ', (1 - len(epochs.events) / len(events)) * 100)
epochs
###################################################################################################
# Run classification
# ----------------------------
clfs = OrderedDict()
clfs['Vect + LR'] = make_pipeline(Vectorizer(), StandardScaler(),
LogisticRegression())
clfs['Vect + RegLDA'] = make_pipeline(Vectorizer(),
LDA(shrinkage='auto', solver='eigen'))
clfs['ERPCov + TS'] = make_pipeline(ERPCovariances(estimator='oas'),
TangentSpace(), LogisticRegression())
clfs['ERPCov + MDM'] = make_pipeline(ERPCovariances(estimator='oas'), MDM())
clfs['XdawnCov + TS'] = make_pipeline(XdawnCovariances(estimator='oas'),
TangentSpace(), LogisticRegression())
clfs['XdawnCov + MDM'] = make_pipeline(XdawnCovariances(estimator='oas'),
MDM())
# format data
epochs.pick_types(eeg=True)
X = epochs.get_data() * 1e6
times = epochs.times
eog=False,
exclude='bads')
epochs = Epochs(raw,
events,
event_id,
tmin,
tmax,
proj=False,
picks=picks,
baseline=None,
preload=True,
verbose=False)
# Create classification pipeline
clf = make_pipeline(Xdawn(n_components=3), Vectorizer(), MinMaxScaler(),
LogisticRegression(penalty='l1'))
# Get the labels
labels = epochs.events[:, -1]
# Cross validator
cv = StratifiedKFold(y=labels, n_folds=10, shuffle=True, random_state=42)
# Do cross-validation
preds = np.empty(len(labels))
for train, test in cv:
clf.fit(epochs[train], labels[train])
preds[test] = clf.predict(epochs[test])
# Classification report
epochs = Epochs(raw,
events,
event_id,
tmin,
tmax,
proj=False,
picks=picks,
baseline=None,
preload=True,
verbose=False)
X = epochs.get_data()
y = label_binarize(epochs.events[:, 2], classes=[1, 3]).ravel()
clf = make_pipeline(XdawnTransformer(n_components=2), Vectorizer(),
StandardScaler(), LogisticRegression())
# Define a monte-carlo cross-validation generator (reduce variance):
cv = ShuffleSplit(len(y), 10, test_size=0.2, random_state=42)
scores = cross_val_score(clf, X, y, cv=cv)
class_balance = np.mean(y == y[0])
class_balance = max(class_balance, 1. - class_balance)
print("Classification accuracy: %f / Chance level: %f" %
(np.mean(scores), class_balance))
###############################################################################
# plot Xdawn patterns estimated on full data for visualization
# Decoding in sensor space using a linear SVM
n_times = len(rt_epochs.times)
from sklearn import preprocessing # noqa
from sklearn.svm import SVC # noqa
from sklearn.pipeline import Pipeline # noqa
from sklearn.model_selection import cross_val_score, ShuffleSplit # noqa
from mne.decoding import Vectorizer, FilterEstimator # noqa
scores_x, scores, std_scores = [], [], []
# don't highpass filter because it's epoched data and the signal length
# is small
filt = FilterEstimator(rt_epochs.info, None, 40, fir_design='firwin')
scaler = preprocessing.StandardScaler()
vectorizer = Vectorizer()
clf = SVC(C=1, kernel='linear')
concat_classifier = Pipeline([('filter', filt), ('vector', vectorizer),
('scaler', scaler), ('svm', clf)])
data_picks = mne.pick_types(rt_epochs.info,
meg='grad',
eeg=False,
eog=True,
stim=False,
exclude=raw.info['bads'])
ax = plt.subplot(111)
ax.set_xlabel('Trials')
ax.set_ylabel('Classification score (% correct)')
ax.set_title('Real-time decoding')
test_epochs = test_epochs.crop(0.2, 0.8)
# Select epochs in [train_epochs]
selects = select_events(train_epochs)
selected_train_epochs = train_epochs[selects]
display(train_epochs, selected_train_epochs, test_epochs)
# Get X and y
train_X, train_y = get_X_y(selected_train_epochs)
test_X, test_y = get_X_y(test_epochs)
# Fit and pred ---------------------------------------------
# Init
clf = make_pipeline(
Vectorizer(), StandardScaler(), pca.PCA(n_components=.95),
svm.SVC(
gamma='scale',
kernel='rbf',
class_weight='balanced',
))
# Fit
print('Fitting')
clf.fit(X=train_X, y=train_y)
# Predict
print('Predicting')
crop_pred_y = clf.predict(X=test_X)
break
design = metadata[predictors]
# # dummy code cue variable
# dummies = pd.get_dummies(design[predictors], drop_first=True)
# design = pd.concat([design.drop(predictors, axis=1), dummies], axis=1)
# design.cue_B = design.cue_B - design.cue_B.unique().mean()
# create design matrix
design = patsy.dmatrix("cue", design, return_type='dataframe')
design = design[['cue[T.B]']]
# 4.2) vectorise channel data for linear regression
# data to be analysed
dat = cues[subj].get_data()
dat = dat[:, :, times_to_use]
Y = Vectorizer().fit_transform(dat)
# 4.3) fit linear model with sklearn's LinearRegression
weights = compute_sample_weight(class_weight='balanced',
y=metadata.cue.to_numpy())
linear_model = LinearRegression(n_jobs=n_jobs, fit_intercept=True)
linear_model.fit(design, Y, sample_weight=weights)
# 4.4) extract the resulting coefficients (i.e., betas)
# extract betas
coefs = get_coef(linear_model, 'coef_')
inter = linear_model.intercept_
# 4.5) extract model r_squared
r2 = r2_score(Y, linear_model.predict(design), multioutput='raw_values')
# save model R-squared
evokeds = []
for sbj in sbjs:
print(sbj)
if sbj == 'VP12': #No REM here
continue
if os.path.exists(os.path.join(save_path, sbj + '.p')):
continue
#sleep_epochs = myload(base_path_data, typ='epoch_preprocessed', sbj=sbj, preload=True) # WAKE!
sleep_epochs = myload(sleep_path_data, typ='epoch_preprocessed', sbj=sbj, preload=True) # SLEEP!
sleep_epochs.event_id = sleep_event_id # event_id remapping. For wake this step works during preprocessing # SLEEP !
sleep_epochs = sleep_epochs.crop(tmin=tmin, tmax=tmax)
X1, y1 = get_Xy_balanced(sleep_epochs, contrast1)
clf = make_pipeline(Vectorizer(), StandardScaler(), LinearModel(LogisticRegression(max_iter = 4000))) #StandardScaler(),
cv = StratifiedKFold(n_splits=2, shuffle=True)
coef_folds = []
for train_idx, test_idx in cv.split(X1, y1):
clf.fit(X1[train_idx], y=y1[train_idx])
#scores1.append(clf.score(X1[test_idx], y=y1[test_idx]))
coef_folds.append(get_coef(clf, attr='patterns_', inverse_transform=True))
coef = np.asarray(coef_folds).mean(0).reshape([173, -1]) #mean folds and reshape
evoked = EvokedArray(coef, sleep_epochs.info, tmin=tmin)
evokeds.append(evoked)
ga = mne.grand_average(evokeds)
#SLEEP
