def runShogunSVRWDKernel(train_xt, train_lt, test_xt, svm_c=1, svr_param=0.1):
"""
serialize svr with string kernels
"""
##################################################
# set up svr
feats_train = construct_features(train_xt)
feats_test = construct_features(test_xt)
max_len = len(train_xt[0])
kernel_wdk = WeightedDegreePositionStringKernel(SIZE, 5)
shifts_vector = np.ones(max_len, dtype=np.int32) * NUMSHIFTS
kernel_wdk.set_shifts(shifts_vector)
########
# set up spectrum
use_sign = False
kernel_spec_1 = WeightedCommWordStringKernel(SIZE, use_sign)
#kernel_spec_2 = WeightedCommWordStringKernel(SIZE, use_sign)
########
# combined kernel
kernel = CombinedKernel()
kernel.append_kernel(kernel_wdk)
kernel.append_kernel(kernel_spec_1)
#kernel.append_kernel(kernel_spec_2)
# init kernel
labels = RegressionLabels(train_lt)
# two svr models: epsilon and nu
svr_epsilon = LibSVR(svm_c, svr_param, kernel, labels, LIBSVR_EPSILON_SVR)
print "Ready to train!"
svr_epsilon.train(feats_train)
#svr_nu=LibSVR(svm_c, svr_param, kernel, labels, LIBSVR_NU_SVR)
#svr_nu.train(feats_train)
# predictions
print "Making predictions!"
kernel.init(feats_train, feats_test)
out1_epsilon = svr_epsilon.apply().get_labels()
out2_epsilon = svr_epsilon.apply(feats_test).get_labels()
#out1_nu=svr_epsilon.apply().get_labels()
#out2_nu=svr_epsilon.apply(feats_test).get_labels()
##################################################
# serialize to file
fEpsilon = open(FNEPSILON, 'w+')
#fNu = open(FNNU, 'w+')
svr_epsilon.save(fEpsilon)
#svr_nu.save(fNu)
fEpsilon.close()
#fNu.close()
##################################################
#return out1_epsilon,out2_epsilon,out1_nu,out2_nu ,kernel
return out1_epsilon, out2_epsilon, kernel
def runShogunSVRWDKernel(train_xt, train_lt, test_xt, svm_c=1, svr_param=0.1):
"""
serialize svr with string kernels
"""
##################################################
# set up svr
feats_train = construct_features(train_xt)
feats_test = construct_features(test_xt)
max_len = len(train_xt[0])
kernel_wdk = WeightedDegreePositionStringKernel(SIZE, 5)
shifts_vector = np.ones(max_len, dtype=np.int32)*NUMSHIFTS
kernel_wdk.set_shifts(shifts_vector)
########
# set up spectrum
use_sign = False
kernel_spec_1 = WeightedCommWordStringKernel(SIZE, use_sign)
#kernel_spec_2 = WeightedCommWordStringKernel(SIZE, use_sign)
########
# combined kernel
kernel = CombinedKernel()
kernel.append_kernel(kernel_wdk)
kernel.append_kernel(kernel_spec_1)
#kernel.append_kernel(kernel_spec_2)
# init kernel
labels = RegressionLabels(train_lt)
# two svr models: epsilon and nu
svr_epsilon=LibSVR(svm_c, svr_param, kernel, labels, LIBSVR_EPSILON_SVR)
print "Ready to train!"
svr_epsilon.train(feats_train)
#svr_nu=LibSVR(svm_c, svr_param, kernel, labels, LIBSVR_NU_SVR)
#svr_nu.train(feats_train)
# predictions
print "Making predictions!"
kernel.init(feats_train, feats_test)
out1_epsilon=svr_epsilon.apply().get_labels()
out2_epsilon=svr_epsilon.apply(feats_test).get_labels()
#out1_nu=svr_epsilon.apply().get_labels()
#out2_nu=svr_epsilon.apply(feats_test).get_labels()
##################################################
# serialize to file
fEpsilon = open(FNEPSILON, 'w+')
#fNu = open(FNNU, 'w+')
svr_epsilon.save(fEpsilon)
#svr_nu.save(fNu)
fEpsilon.close()
#fNu.close()
##################################################
#return out1_epsilon,out2_epsilon,out1_nu,out2_nu ,kernel
return out1_epsilon,out2_epsilon,kernel
def runShogunSVMDNAWDKernel(train_xt, train_lt, test_xt):
"""
run svm with string kernels
"""
##################################################
# set up svm
feats_train = StringCharFeatures(train_xt, DNA)
feats_test = StringCharFeatures(test_xt, DNA)
kernel = WeightedDegreePositionStringKernel(feats_train, feats_train,
DEGREE)
kernel.io.set_loglevel(MSG_DEBUG)
kernel.set_shifts(NUMSHIFTS * ones(len(train_xt[0]), dtype=int32))
kernel.set_position_weights(ones(len(train_xt[0]), dtype=float64))
# init kernel
labels = BinaryLabels(train_lt)
# run svm model
print "Ready to train!"
svm = LibSVM(SVMC, kernel, labels)
svm.io.set_loglevel(MSG_DEBUG)
svm.train()
# predictions
print "Making predictions!"
out1DecisionValues = svm.apply(feats_train)
out1 = out1DecisionValues.get_labels()
kernel.init(feats_train, feats_test)
out2DecisionValues = svm.apply(feats_test)
out2 = out2DecisionValues.get_labels()
return out1, out2, out1DecisionValues, out2DecisionValues
def kernel_weighted_degree_position_string_modular(fm_train_dna=traindat,
fm_test_dna=testdat,
degree=20):
from modshogun import StringCharFeatures, DNA
from modshogun import WeightedDegreePositionStringKernel, MSG_DEBUG
feats_train = StringCharFeatures(fm_train_dna, DNA)
#feats_train.io.set_loglevel(MSG_DEBUG)
feats_test = StringCharFeatures(fm_test_dna, DNA)
kernel = WeightedDegreePositionStringKernel(feats_train, feats_train,
degree)
from numpy import zeros, ones, float64, int32
kernel.set_shifts(10 * ones(len(fm_train_dna[0]), dtype=int32))
kernel.set_position_weights(ones(len(fm_train_dna[0]), dtype=float64))
km_train = kernel.get_kernel_matrix()
kernel.init(feats_train, feats_test)
km_test = kernel.get_kernel_matrix()
return km_train, km_test, kernel
def kernel_weighted_degree_position_string_modular (fm_train_dna=traindat,fm_test_dna=testdat,degree=20): from modshogun import StringCharFeatures, DNA from modshogun import WeightedDegreePositionStringKernel, MSG_DEBUG feats_train=StringCharFeatures(fm_train_dna, DNA) #feats_train.io.set_loglevel(MSG_DEBUG) feats_test=StringCharFeatures(fm_test_dna, DNA) kernel=WeightedDegreePositionStringKernel(feats_train, feats_train, degree) from numpy import zeros,ones,float64,int32 kernel.set_shifts(10*ones(len(fm_train_dna[0]), dtype=int32)) kernel.set_position_weights(ones(len(fm_train_dna[0]), dtype=float64)) km_train=kernel.get_kernel_matrix() kernel.init(feats_train, feats_test) km_test=kernel.get_kernel_matrix() return km_train,km_test,kernel
def svm_process(args_tuple):
X_train, Y_train, X_test, Y_test, d, c = args_tuple
kernel = WeightedDegreePositionStringKernel(X_train, X_train, d)
kernel.set_shifts(np.ones(SEQ_LEN, dtype=np.int32))
kernel.set_position_weights(np.ones(SEQ_LEN, dtype=np.float64))
kernel.init(X_train, X_train)
model = SVMLight(c, kernel, Y_train)
model.train()
Y_test_pred = model.apply(X_test).get_labels()
Y_test_dist = model.apply(X_test).get_values()
Y_test_proba = np.exp(Y_test_dist) / (1 + np.exp(Y_test_dist))
accuracy = np.where(Y_test_pred - Y_test == 0)[0].size * 1.0 / Y_test.size
return (accuracy, Y_test_proba)
def svm_process(args_tuple):
X_train, Y_train, X_test, Y_test, d, c = args_tuple
kernel = WeightedDegreePositionStringKernel(X_train, X_train, d)
kernel.set_shifts(np.ones(SEQ_LEN, dtype=np.int32))
kernel.set_position_weights(np.ones(SEQ_LEN, dtype=np.float64))
kernel.init(X_train, X_train)
model = SVMLight(c, kernel, Y_train)
model.train()
Y_test_pred = model.apply(X_test).get_labels()
Y_test_dist = model.apply(X_test).get_values()
Y_test_proba = np.exp(Y_test_dist)/(1 + np.exp(Y_test_dist))
accuracy = np.where(Y_test_pred - Y_test == 0)[0].size*1.0/Y_test.size
return (accuracy, Y_test_proba)
def runShogunSVRWDKernel(train_xt, train_lt, test_xt, svm_c=1):
"""
serialize svr with string kernels
"""
##################################################
# set up svr
feats_train = StringCharFeatures(train_xt, PROTEIN)
feats_test = StringCharFeatures(test_xt, PROTEIN)
kernel = WeightedDegreePositionStringKernel(feats_train, feats_train, DEGREE)
kernel.io.set_loglevel(MSG_DEBUG)
kernel.set_shifts(NUMSHIFTS*ones(len(train_xt[0]), dtype=int32))
kernel.set_position_weights(ones(len(train_xt[0]), dtype=float64))
# init kernel
labels = RegressionLabels(train_lt)
# two svr models: epsilon and nu
print "Ready to train!"
svr_epsilon=LibSVR(svm_c, SVRPARAM, kernel, labels, LIBSVR_EPSILON_SVR)
svr_epsilon.io.set_loglevel(MSG_DEBUG)
svr_epsilon.train()
#svr_nu=LibSVR(svm_c, SVRPARAM, kernel, labels, LIBSVR_NU_SVR)
#svr_nu.train()
# predictions
print "Making predictions!"
out1_epsilon=svr_epsilon.apply(feats_train).get_labels()
kernel.init(feats_train, feats_test)
out2_epsilon=svr_epsilon.apply(feats_test).get_labels()
#out1_nu=svr_epsilon.apply(feats_train).get_labels()
#out2_nu=svr_epsilon.apply(feats_test).get_labels()
#return out1_epsilon,out2_epsilon,out1_nu,out2_nu ,kernel
return out1_epsilon,out2_epsilon,kernel
def serialization_string_kernels_modular(n_data, num_shifts, size):
"""
serialize svm with string kernels
"""
##################################################
# set up toy data and svm
train_xt, train_lt = generate_random_data(n_data)
test_xt, test_lt = generate_random_data(n_data)
feats_train = construct_features(train_xt)
feats_test = construct_features(test_xt)
max_len = len(train_xt[0])
kernel_wdk = WeightedDegreePositionStringKernel(size, 5)
shifts_vector = numpy.ones(max_len, dtype=numpy.int32) * num_shifts
kernel_wdk.set_shifts(shifts_vector)
########
# set up spectrum
use_sign = False
kernel_spec_1 = WeightedCommWordStringKernel(size, use_sign)
kernel_spec_2 = WeightedCommWordStringKernel(size, use_sign)
########
# combined kernel
kernel = CombinedKernel()
kernel.append_kernel(kernel_wdk)
kernel.append_kernel(kernel_spec_1)
kernel.append_kernel(kernel_spec_2)
# init kernel
labels = BinaryLabels(train_lt)
svm = SVMLight(1.0, kernel, labels)
#svm.io.set_loglevel(MSG_DEBUG)
svm.train(feats_train)
##################################################
# serialize to file
fn = "serialized_svm.bz2"
#print("serializing SVM to file", fn)
save(fn, svm)
##################################################
# unserialize and sanity check
#print("unserializing SVM")
svm2 = load(fn)
#print("comparing predictions")
out = svm.apply(feats_test).get_labels()
out2 = svm2.apply(feats_test).get_labels()
# assert outputs are close
for i in range(len(out)):
assert abs(out[i] - out2[i] < 0.000001)
#print("all checks passed.")
return out, out2
def serialization_string_kernels_modular(n_data, num_shifts, size):
"""
serialize svm with string kernels
"""
##################################################
# set up toy data and svm
train_xt, train_lt = generate_random_data(n_data)
test_xt, test_lt = generate_random_data(n_data)
feats_train = construct_features(train_xt)
feats_test = construct_features(test_xt)
max_len = len(train_xt[0])
kernel_wdk = WeightedDegreePositionStringKernel(size, 5)
shifts_vector = numpy.ones(max_len, dtype=numpy.int32)*num_shifts
kernel_wdk.set_shifts(shifts_vector)
########
# set up spectrum
use_sign = False
kernel_spec_1 = WeightedCommWordStringKernel(size, use_sign)
kernel_spec_2 = WeightedCommWordStringKernel(size, use_sign)
########
# combined kernel
kernel = CombinedKernel()
kernel.append_kernel(kernel_wdk)
kernel.append_kernel(kernel_spec_1)
kernel.append_kernel(kernel_spec_2)
# init kernel
labels = BinaryLabels(train_lt);
svm = SVMLight(1.0, kernel, labels)
#svm.io.set_loglevel(MSG_DEBUG)
svm.train(feats_train)
##################################################
# serialize to file
fn = "serialized_svm.bz2"
#print("serializing SVM to file", fn)
save(fn, svm)
##################################################
# unserialize and sanity check
#print("unserializing SVM")
svm2 = load(fn)
#print("comparing predictions")
out = svm.apply(feats_test).get_labels()
out2 = svm2.apply(feats_test).get_labels()
# assert outputs are close
for i in range(len(out)):
assert abs(out[i] - out2[i] < 0.000001)
#print("all checks passed.")
return out,out2