def run(fname, folds):
st = time.clock()
data = get_data("Data/{}.data".format(fname))
with open("Results/{}_{}folds.txt".format(fname, folds), "a") as f:
tee = Tee(sys.stdout, f)
tee("Validating classifier with {}-fold test...".format(folds))
kf = KFold(len(data), n_folds=folds)
avg_error = 0
it = 1
for train, test in kf:
start = time.clock()
tee("Iteration #{}".format(it))
oc = ObliqueClassifier()
oc.fit(data[train])
predictions = [oc.predict(r) for r in data[test]]
actual_labels = data[test][:, -1]
error = error_rate(predictions, actual_labels)
tee("Error: {:.3f}".format(error))
tee("Elapsed time: {:.3f} seconds".format(time.clock() - start))
tee()
avg_error += error
it += 1
totaltime = time.clock() - st
tee("Average error: {:.3f}".format(avg_error/folds))
tee("Total elapsed time: {:.3f} seconds.".format(totaltime))
tee("Average elapsed time: {:.3f} seconds.".format(totaltime/folds))
def run(fname, folds):
st = time.clock()
data = get_data("Data/{}.data".format(fname))
with open("Results/{}_{}folds.txt".format(fname, folds), "a") as f:
tee = Tee(sys.stdout, f)
tee("Validating classifier with {}-fold test...".format(folds))
kf = KFold(len(data), n_folds=folds)
avg_error = 0
it = 1
for train, test in kf:
start = time.clock()
tee("Iteration #{}".format(it))
oc = ObliqueClassifier()
oc.fit(data[train])
predictions = [oc.predict(r) for r in data[test]]
actual_labels = data[test][:, -1]
error = error_rate(predictions, actual_labels)
tee("Error: {:.3f}".format(error))
tee("Elapsed time: {:.3f} seconds".format(time.clock() - start))
tee()
avg_error += error
it += 1
totaltime = time.clock() - st
tee("Average error: {:.3f}".format(avg_error / folds))
tee("Total elapsed time: {:.3f} seconds.".format(totaltime))
tee("Average elapsed time: {:.3f} seconds.".format(totaltime / folds))
#!/usr/bin/env python
# This little transformation makes the dataset a lot smaller
# but the classifier still takes absolute ages to fit it.
from pyblique import get_data
from sklearn.decomposition import PCA
import numpy as np
# Shrink the isolet dataset by means of PCA.
# 600+ attributes -> 10 attributes
isolet = get_data("Data/isolet.data")
pca = PCA(n_components=10)
transformed = pca.fit_transform(isolet[:, :-1])
labels = isolet[:, -1]
labels = np.reshape(labels, (labels.shape[0], 1))
result = np.concatenate((transformed, labels), axis=1)
np.savetxt("Data/isolet_compressed.data", result, delimiter=",")
#!/usr/bin/env python
# This little transformation makes the dataset a lot smaller
# but the classifier still takes absolute ages to fit it.
from pyblique import get_data
from sklearn.decomposition import PCA
import numpy as np
# Shrink the isolet dataset by means of PCA.
# 600+ attributes -> 10 attributes
isolet = get_data("Data/isolet.data")
pca = PCA(n_components=10)
transformed = pca.fit_transform(isolet[:, :-1])
labels = isolet[:, -1]
labels = np.reshape(labels, (labels.shape[0], 1))
result = np.concatenate((transformed, labels), axis=1)
np.savetxt("Data/isolet_compressed.data", result, delimiter=",")
