def nfoldCV_SVM(Features, Labels, KF, C, i, Scores, kernel):
nIterate = 1
score = np.zeros([10])
jobs = []
op.colPrint("C: %s" % C, col="c")
for train, test in KF:
jobs.append(
threading.Thread(target=one_trial_SVM, args=(nIterate, Features, Labels, train, test, C, score, kernel))
)
nIterate = nIterate + 1
for j in jobs:
j.start()
for j in jobs:
j.join()
Scores[i, :] = score
op.colPrint("C: %s; done." % C, col="y")
import sys
from sklearn import svm
from sklearn.externals import joblib
import option_10foldcv as op
if __name__ == '__main__':
if len(sys.argv) < 2:
modelName = "clf.model"
else:
modelName = sys.argv[1]
# read feature vector and corresponding labels ======================
op.colPrint("Reading features & labels ================", col='c')
Features = op.readFeatures()
Labels = op.readLabels()
if Features.shape[0] != Labels.shape[0]:
sys.exit("The number of samples and labels are different.")
nSamples = Features.shape[0]
op.colPrint("Reading features & labels; done. Number of samples: %s" % nSamples, col='y')
# create svm instance and training
clf = svm.SVC(kernel='linear', C=10**1, probability=True).fit(Features, Labels)
# export classifier
print "Save classifier as", modelName
joblib.dump(clf, modelName)
# main ==================================================================
if __name__ == "__main__":
pid = os.getpid()
print "pid:", pid
if len(sys.argv) < 2:
kernel = "linear"
else:
kernel = sys.argv[1]
print "kernel:", kernel
# read feature vector and corresponding labels ======================
op.colPrint("Reading features & labels ================", col="c")
Features = op.readFeatures()
Labels = op.readLabels()
if Features.shape[0] != Labels.shape[0]:
sys.exit("The number of samples and labels are different.")
nSamples = Features.shape[0]
op.colPrint("Reading features & labels; done. Number of samples: %s" % nSamples, col="y")
# split samples (N = 10) ============================================
kf = KFold(nSamples, n_folds=10, shuffle=True, random_state=None)
# Parameter of SVM ===========================
C = np.logspace(-5, 5, num=11)
# C = np.logspace(-5, 4, num=10)
Scores = np.zeros([len(C), 10])
