def compare_by_subject(tasks, subjects, position, sessionnum=""):
for i in range(len(tasks)):
print('For task ' + tasks[i] + ', position ' + str(position) + ':')
for j in range(len(subjects)):
for k in range(j+1, len(subjects)):
subA = feature_vector_generator(tasks[i], subjects[j], position, sessionnum)
subB = feature_vector_generator(tasks[i], subjects[k], position, sessionnum)
X, y = labeler.vectorsAndLabels([subA, subB])
comparison = 'subject' + str(subjects[j]) + " vs. subject" + str(subjects[k]) + " cross-validation is: " + str(labeler.crossValidate(X, y))
print(comparison)
def compare_by_task(tasks, subject1, position, sessionnum="", sq=""):
print('For subject ' + str(subject1) + ', position ' + str(position) + ':')
for i in range(len(tasks)):
for j in range(i+1, len(tasks)):
for k in range(j+1, len(tasks)):
for l in range(k+1, len(tasks)):
taskA = feature_vector_generator(tasks[i], subject1, position, sessionnum, sq)
taskB = feature_vector_generator(tasks[j], subject1, position, sessionnum, sq)
taskC = feature_vector_generator(tasks[k], subject1, position, sessionnum, sq)
taskD = feature_vector_generator(tasks[l], subject1, position, sessionnum, sq)
X, y = labeler.vectorsAndLabels([taskA, taskB, taskC, taskD])
comparison = tasks[i] + " vs. " + tasks[j] + " vs. " + tasks[k] + " vs. " + tasks[l] + " cross-validation is: " + str(labeler.crossValidate(X, y))
print(comparison)
def num_feature_vectors(subjects, t0, t1):
def genlength(gen):
return len([n for n in gen])
return [{
subject: genlength(feature_vector_generator(subject, t0, t1))
} for subject in subjects]
def compare_across_all_tasks(tasks, subject, position, sessionnum="", sq=""):
print('For subject ' + str(subject) + ':')
for i in range(len(tasks)):
taskX = feature_vector_generator(tasks[i], subject, position, sessionnum, sq)
def f():
return
yield
g = f()
for j in range(i):
# print(tasks[j])
taskA = feature_vector_generator(tasks[j], subject, position, sessionnum, sq)
g = chain(g, taskA)
for k in range(i+1, len(tasks)):
# print(tasks[k])
taskB = feature_vector_generator(tasks[k], subject, position, sessionnum, sq)
g = chain(g, taskB)
X, y = labeler.vectorsAndLabels([taskX, g])
comparison = tasks[i] + " vs. others cross-validation is: " + str(labeler.crossValidate(X, y))
print(comparison)
def make_svm_classifier(tasks, subject, position, sessionnum="", sq=""):
for i in range(len(tasks)):
print('For subject ' + str(subject) + ', task ' + str(tasks[i]) + ':')
taskX = feature_vector_generator(tasks[i], subject, position, sessionnum, sq)
def f():
return
yield
g = f()
for j in range(i):
# print(tasks[j])
taskA = feature_vector_generator(tasks[j], subject, position, sessionnum, sq)
g = chain(g, taskA)
for k in range(i+1, len(tasks)):
# print(tasks[k])
taskB = feature_vector_generator(tasks[k], subject, position, sessionnum, sq)
g = chain(g, taskB)
X, y = labeler.vectorsAndLabels([taskX, g])
'''Toggle the following for learner function or cross-validation.'''
# lin_clf = LinearSVC()
# target = lin_clf.fit(X, y)
# return target
comparison = tasks[i] + " vs. others cross-validation is: " + str(labeler.crossValidate(X, y))
print(comparison)
def num_feature_vectors(subjects, t0, t1):
def genlength(gen):
return len([n for n in gen])
return [{subject: genlength(feature_vector_generator(subject, t0, t1))}
for subject in subjects]
def vectorsAndLabels(arrayOfGenerators):
'''Takes an array of generators and produces two lists X and y
where len(X) = len(y),
X is the vectors, y is their (numerical) labels.'''
X = []
y = []
currentLabel = 0
for generator in arrayOfGenerators:
for vector in generator:
# NB: feature vectors are serialized as strings sometimes
# we map them to floats
X.append(vector)
y.append(currentLabel)
currentLabel += 1
return X, y
def crossValidate(X, y):
"7-fold cross-validation with an SVM with a set of labels and vectors"
clf = svm.LinearSVC()
scores = cross_validation.cross_val_score(clf, np.array(X), y, cv=7)
return scores.mean()
personA_gen = feature_vector_generator('cube', 2, 2)
personB_gen = feature_vector_generator('cube', 1, 2)
X, y = vectorsAndLabels([personA_gen, personB_gen])
print(crossValidate(X, y))
y = []
currentLabel = 0
for generator in arrayOfGenerators:
for vector in generator:
# NB: feature vectors are serialized as strings sometimes
# we map them to floats
X.append(vector)
y.append(currentLabel)
currentLabel += 1
return X, y
def crossValidate(X, y):
"7-fold cross-validation with an SVM with a set of labels and vectors"
clf = svm.LinearSVC()
scores = cross_validation.cross_val_score(clf, np.array(X), y, cv=7)
return scores.mean()
# let's see how well we can distinguish between two subjects based on their brainwaves.
# we'll get their data from a specific time range:
t0 = parse('2015-05-09 23:28:00+00')
t1 = parse('2015-05-09 23:30:31+00')
# and make two generators of feature vectors for the two different subjects:
personA_gen = feature_vector_generator(9, t0, t1)
personB_gen = feature_vector_generator(13, t0, t1)
# now let's feed these feature vectors into an SVM
# and do 7-fold cross-validation.
X, y = vectorsAndLabels([personA_gen, personB_gen])
print crossValidate(X, y)
X is the vectors, y is their (numerical) labels.'''
X = []
y = []
currentLabel = 0
for generator in arrayOfGenerators:
for vector in generator:
# NB: feature vectors are serialized as strings sometimes
# we map them to floats
X.append(vector)
y.append(currentLabel)
currentLabel+=1
return X, y
def crossValidate(X,y):
"7-fold cross-validation with an SVM with a set of labels and vectors"
clf = svm.LinearSVC()
scores = cross_validation.cross_val_score(clf, np.array(X), y, cv=7)
return scores.mean()
# let's see how well we can distinguish between two subjects based on their brainwaves.
# we'll get their data from a specific time range:
t0 = parse('2015-05-09 23:28:00+00')
t1 = parse('2015-05-09 23:30:31+00')
# and make two generators of feature vectors for the two different subjects:
personA_gen = feature_vector_generator(9, t0, t1)
personB_gen = feature_vector_generator(13, t0, t1)
# now let's feed these feature vectors into an SVM
# and do 7-fold cross-validation.
X, y = vectorsAndLabels([personA_gen, personB_gen])
print crossValidate(X, y)
