def evaluation_thresholds_modular(index):
from modshogun import BinaryLabels, ROCEvaluation
import numpy
numpy.random.seed(17)
output = numpy.arange(-1, 1, 0.001)
output = (0.3 * output + 0.7 * (numpy.random.rand(len(output)) - 0.5))
label = [-1.0] * (len(output) // 2)
label.extend([1.0] * (len(output) // 2))
label = numpy.array(label)
pred = BinaryLabels(output)
truth = BinaryLabels(label)
evaluator = ROCEvaluation()
evaluator.evaluate(pred, truth)
[fp, tp] = evaluator.get_ROC()
thresh = evaluator.get_thresholds()
b = thresh[index]
#print("tpr", numpy.mean(output[label>0]>b), tp[index])
#print("fpr", numpy.mean(output[label<0]>b), fp[index])
return tp[index], fp[index], numpy.mean(output[label > 0] > b), numpy.mean(
output[label < 0] > b)
def classifier_domainadaptationsvm_modular (fm_train_dna=traindna,fm_test_dna=testdna, \
label_train_dna=label_traindna, \
label_test_dna=label_testdna,fm_train_dna2=traindna2,fm_test_dna2=testdna2, \
label_train_dna2=label_traindna2,label_test_dna2=label_testdna2,C=1,degree=3):
feats_train = StringCharFeatures(fm_train_dna, DNA)
feats_test = StringCharFeatures(fm_test_dna, DNA)
kernel = WeightedDegreeStringKernel(feats_train, feats_train, degree)
labels = BinaryLabels(label_train_dna)
svm = SVMLight(C, kernel, labels)
svm.train()
#svm.io.set_loglevel(MSG_DEBUG)
#####################################
#print("obtaining DA SVM from previously trained SVM")
feats_train2 = StringCharFeatures(fm_train_dna, DNA)
feats_test2 = StringCharFeatures(fm_test_dna, DNA)
kernel2 = WeightedDegreeStringKernel(feats_train, feats_train, degree)
labels2 = BinaryLabels(label_train_dna)
# we regularize against the previously obtained solution
dasvm = DomainAdaptationSVM(C, kernel2, labels2, svm, 1.0)
dasvm.train()
out = dasvm.apply_binary(feats_test2)
return out #,dasvm TODO
def evaluation_director_contingencytableevaluation_modular(
ground_truth, predicted):
try:
from modshogun import DirectorContingencyTableEvaluation, ED_MAXIMIZE
except ImportError:
print("recompile shogun with --enable-swig-directors")
return
class SimpleWeightedBinaryEvaluator(DirectorContingencyTableEvaluation):
def __init__(self):
DirectorContingencyTableEvaluation.__init__(self)
def get_custom_direction(self):
return ED_MAXIMIZE
def get_custom_score(self):
return self.get_WRACC() + self.get_BAL()
from modshogun import BinaryLabels
evaluator = SimpleWeightedBinaryEvaluator()
r = evaluator.evaluate(BinaryLabels(ground_truth), BinaryLabels(predicted))
r2 = evaluator.get_custom_score()
print(r, r2)
return r, r2
def outputResultsClassificationWithMajorityClass(out1, out2, out1DecisionValues, out2DecisionValues, train_lt, test_lt, test_majorityClass): # Output the results to the appropriate output files writeFloatList(out1, TRAINPREDICTIONSEPSILONFILENAME) writeFloatList(out2, VALIDATIONPREDICTIONSEPSILONFILENAME) numTrainCorrect = 0 for i in range(len(train_lt)): # Iterate through training labels and count the number that are the same as the predicted labels if out1[i] == train_lt[i]: # The current prediction is correct numTrainCorrect = numTrainCorrect + 1 fracTrainCorrect = float(numTrainCorrect)/float(len(train_lt)) print "Training accuracy:" print fracTrainCorrect trainLabels = BinaryLabels(train_lt) evaluatorTrain = ROCEvaluation() evaluatorTrain.evaluate(out1DecisionValues, trainLabels) print "Training AUC:" print evaluatorTrain.get_auROC() numValidCorrect = 0 numPosCorrect = 0 numNegCorrect = 0 numMajorityClassCorrect = 0 numMinorityClassCorrect = 0 for i in range(len(test_lt)): # Iterate through validation labels and count the number that are the same as the predicted labels if out2[i] == test_lt[i]: # The current prediction is correct numValidCorrect = numValidCorrect + 1 if (out2[i] == 1) and (test_lt[i] == 1): # The prediction is a positive example numPosCorrect = numPosCorrect + 1 else: numNegCorrect = numNegCorrect + 1 if test_majorityClass[i] == 1: numMajorityClassCorrect = numMajorityClassCorrect + 1 else: numMinorityClassCorrect = numMinorityClassCorrect + 1 fracValidCorrect = float(numValidCorrect)/float(len(test_lt)) print "Validation accuracy:" print fracValidCorrect print "Fraction of correct positive examples:" print float(numPosCorrect)/float(len(np.where(test_lt > 0)[0])) print "Fraction of correct negative examples:" print float(numNegCorrect)/float(len(np.where(test_lt <= 0)[0])) print "Fraction of correct majority class examples:" print float(numMajorityClassCorrect)/float(len(np.where(test_majorityClass > 0)[0])) print "Fraction of correct minority class examples:" print float(numMinorityClassCorrect)/float(len(np.where(test_majorityClass <= 0)[0])) validLabels = BinaryLabels(test_lt) evaluatorValid = ROCEvaluation() evaluatorValid.evaluate(out2DecisionValues, validLabels) print "Validation AUC:" print evaluatorValid.get_auROC()
def modelselection_grid_search_kernel(num_subsets, num_vectors, dim_vectors):
# init seed for reproducability
Math.init_random(1)
random.seed(1)
# create some (non-sense) data
matrix = random.rand(dim_vectors, num_vectors)
# create num_feautres 2-dimensional vectors
features = RealFeatures()
features.set_feature_matrix(matrix)
# create labels, two classes
labels = BinaryLabels(num_vectors)
for i in range(num_vectors):
labels.set_label(i, 1 if i % 2 == 0 else -1)
# create svm
classifier = LibSVM()
# splitting strategy
splitting_strategy = StratifiedCrossValidationSplitting(
labels, num_subsets)
# accuracy evaluation
evaluation_criterion = ContingencyTableEvaluation(ACCURACY)
# cross validation class for evaluation in model selection
cross = CrossValidation(classifier, features, labels, splitting_strategy,
evaluation_criterion)
cross.set_num_runs(1)
# print all parameter available for modelselection
# Dont worry if yours is not included, simply write to the mailing list
#classifier.print_modsel_params()
# model parameter selection
param_tree = create_param_tree()
#param_tree.print_tree()
grid_search = GridSearchModelSelection(cross, param_tree)
print_state = False
best_combination = grid_search.select_model(print_state)
#print("best parameter(s):")
#best_combination.print_tree()
best_combination.apply_to_machine(classifier)
# larger number of runs to have tighter confidence intervals
cross.set_num_runs(10)
cross.set_conf_int_alpha(0.01)
result = cross.evaluate()
casted = CrossValidationResult.obtain_from_generic(result)
#print "result mean:", casted.mean
return classifier, result, casted.mean
def modelselection_grid_search_kernel (num_subsets, num_vectors, dim_vectors):
# init seed for reproducability
Math.init_random(1)
random.seed(1);
# create some (non-sense) data
matrix=random.rand(dim_vectors, num_vectors)
# create num_feautres 2-dimensional vectors
features=RealFeatures()
features.set_feature_matrix(matrix)
# create labels, two classes
labels=BinaryLabels(num_vectors)
for i in range(num_vectors):
labels.set_label(i, 1 if i%2==0 else -1)
# create svm
classifier=LibSVM()
# splitting strategy
splitting_strategy=StratifiedCrossValidationSplitting(labels, num_subsets)
# accuracy evaluation
evaluation_criterion=ContingencyTableEvaluation(ACCURACY)
# cross validation class for evaluation in model selection
cross=CrossValidation(classifier, features, labels, splitting_strategy, evaluation_criterion)
cross.set_num_runs(1)
# print all parameter available for modelselection
# Dont worry if yours is not included, simply write to the mailing list
#classifier.print_modsel_params()
# model parameter selection
param_tree=create_param_tree()
#param_tree.print_tree()
grid_search=GridSearchModelSelection(cross, param_tree)
print_state=False
best_combination=grid_search.select_model(print_state)
#print("best parameter(s):")
#best_combination.print_tree()
best_combination.apply_to_machine(classifier)
# larger number of runs to have tighter confidence intervals
cross.set_num_runs(10)
cross.set_conf_int_alpha(0.01)
result=cross.evaluate()
casted=CrossValidationResult.obtain_from_generic(result);
#print "result mean:", casted.mean
return classifier,result,casted.mean
def evaluation_rocevaluation_modular(ground_truth, predicted):
from modshogun import BinaryLabels
from modshogun import ROCEvaluation
ground_truth_labels = BinaryLabels(ground_truth)
predicted_labels = BinaryLabels(predicted)
evaluator = ROCEvaluation()
evaluator.evaluate(predicted_labels, ground_truth_labels)
return evaluator.get_ROC(), evaluator.get_auROC()
def label_function():
from modshogun import BinaryLabels
from modshogun import CSVFile
#generate random labels
label = BinaryLabels(5)
print label.get_num_labels()
#→ 5
print label.get_values()
#→ array([5 label values])
#Labels from CSVFile
label_from_csv = BinaryLabels(CSVFile("csv/label.csv"))
def classifier_featureblock_logistic_regression(fm_train=traindat,
fm_test=testdat,
label_train=label_traindat):
from modshogun import BinaryLabels, RealFeatures, IndexBlock, IndexBlockGroup
try:
from modshogun import FeatureBlockLogisticRegression
except ImportError:
print("FeatureBlockLogisticRegression not available")
exit(0)
features = RealFeatures(hstack((traindat, traindat)))
labels = BinaryLabels(hstack((label_train, label_train)))
n_features = features.get_num_features()
block_one = IndexBlock(0, n_features // 2)
block_two = IndexBlock(n_features // 2, n_features)
block_group = IndexBlockGroup()
block_group.add_block(block_one)
block_group.add_block(block_two)
mtlr = FeatureBlockLogisticRegression(0.1, features, labels, block_group)
mtlr.set_regularization(1) # use regularization ratio
mtlr.set_tolerance(1e-2) # use 1e-2 tolerance
mtlr.train()
out = mtlr.apply().get_labels()
return out
def cross_validation(X, Y, d, c, K):
N = len(Y)
n = N / K
accuracy_list = []
for k in range(0, K):
print 'degree = %s\tC = %s\tcross_validation_iter = %s/%s' % (d, c,
k + 1, K)
sys.stdout.flush()
X_test = list(X[k:k + n])
Y_test = list(Y[k:k + n])
X_train = []
X_train.extend(X[:k])
X_train.extend(X[k + n:])
Y_train = []
Y_train.extend(Y[:k])
Y_train.extend(Y[k + n:])
X_train = StringCharFeatures(X_train, DNA)
X_test = StringCharFeatures(X_test, DNA)
Y_train = BinaryLabels(np.array(Y_train, dtype=np.float64))
Y_test = np.array(Y_test)
args_tuple = (X_train, Y_train, X_test, Y_test, d, c)
accuracy, Y_test_proba = svm_process(args_tuple)
accuracy_list.append(accuracy)
return np.array(accuracy_list).mean()
def classifier_gpbtsvm_modular(train_fname=traindat,
test_fname=testdat,
label_fname=label_traindat,
width=2.1,
C=1,
epsilon=1e-5):
from modshogun import RealFeatures, BinaryLabels
from modshogun import GaussianKernel
from modshogun import CSVFile
try:
from modshogun import GPBTSVM
except ImportError:
print("GPBTSVM not available")
exit(0)
feats_train = RealFeatures(CSVFile(train_fname))
feats_test = RealFeatures(CSVFile(test_fname))
labels = BinaryLabels(CSVFile(label_fname))
kernel = GaussianKernel(feats_train, feats_train, width)
svm = GPBTSVM(C, kernel, labels)
svm.set_epsilon(epsilon)
svm.train()
predictions = svm.apply(feats_test)
return predictions, svm, predictions.get_labels()
def transfer_multitask_l12_logistic_regression(fm_train=traindat,
fm_test=testdat,
label_train=label_traindat):
from modshogun import BinaryLabels, RealFeatures, Task, TaskGroup
try:
from modshogun import MultitaskL12LogisticRegression
except ImportError:
print("MultitaskL12LogisticRegression not available")
exit(0)
features = RealFeatures(hstack((traindat, traindat)))
labels = BinaryLabels(hstack((label_train, label_train)))
n_vectors = features.get_num_vectors()
task_one = Task(0, n_vectors // 2)
task_two = Task(n_vectors // 2, n_vectors)
task_group = TaskGroup()
task_group.append_task(task_one)
task_group.append_task(task_two)
mtlr = MultitaskL12LogisticRegression(0.1, 0.1, features, labels,
task_group)
mtlr.set_tolerance(1e-2) # use 1e-2 tolerance
mtlr.set_max_iter(10)
mtlr.train()
mtlr.set_current_task(0)
out = mtlr.apply_regression().get_labels()
return out
def classifier_svmlight_linear_term_modular (fm_train_dna=traindna,fm_test_dna=testdna, \
label_train_dna=label_traindna,degree=3, \
C=10,epsilon=1e-5,num_threads=1):
from modshogun import StringCharFeatures, BinaryLabels, DNA
from modshogun import WeightedDegreeStringKernel
try:
from modshogun import SVMLight
except ImportError:
print("SVMLight is not available")
exit(0)
feats_train = StringCharFeatures(DNA)
feats_train.set_features(fm_train_dna)
feats_test = StringCharFeatures(DNA)
feats_test.set_features(fm_test_dna)
kernel = WeightedDegreeStringKernel(feats_train, feats_train, degree)
labels = BinaryLabels(label_train_dna)
svm = SVMLight(C, kernel, labels)
svm.set_qpsize(3)
svm.set_linear_term(
-numpy.array([1, 2, 3, 4, 5, 6, 7, 8, 7, 6], dtype=numpy.double))
svm.set_epsilon(epsilon)
svm.parallel.set_num_threads(num_threads)
svm.train()
kernel.init(feats_train, feats_test)
out = svm.apply().get_labels()
return out, 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 runShogunSVMDNASubsequenceStringKernel(train_xt, train_lt, test_xt):
"""
run svm with spectrum kernel
"""
##################################################
# set up svm
feats_train = StringCharFeatures(train_xt, DNA)
feats_test = StringCharFeatures(test_xt, DNA)
kernel = SubsequenceStringKernel(feats_train, feats_train, MAXLEN, DECAY)
kernel.io.set_loglevel(MSG_DEBUG)
kernel.init(feats_train, feats_train)
# 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 transfer_multitask_clustered_logistic_regression(fm_train=traindat,
fm_test=testdat,
label_train=label_traindat
):
from modshogun import BinaryLabels, RealFeatures, Task, TaskGroup, MultitaskClusteredLogisticRegression, MSG_DEBUG
features = RealFeatures(hstack((traindat, sin(traindat), cos(traindat))))
labels = BinaryLabels(hstack((label_train, label_train, label_train)))
n_vectors = features.get_num_vectors()
task_one = Task(0, n_vectors // 3)
task_two = Task(n_vectors // 3, 2 * n_vectors // 3)
task_three = Task(2 * n_vectors // 3, n_vectors)
task_group = TaskGroup()
task_group.append_task(task_one)
task_group.append_task(task_two)
task_group.append_task(task_three)
mtlr = MultitaskClusteredLogisticRegression(1.0, 100.0, features, labels,
task_group, 2)
#mtlr.io.set_loglevel(MSG_DEBUG)
mtlr.set_tolerance(1e-3) # use 1e-2 tolerance
mtlr.set_max_iter(100)
mtlr.train()
mtlr.set_current_task(0)
#print mtlr.get_w()
out = mtlr.apply_regression().get_labels()
return out
def classifier_svmlight_modular(fm_train_dna=traindat,
fm_test_dna=testdat,
label_train_dna=label_traindat,
C=1.2,
epsilon=1e-5,
num_threads=1):
from modshogun import StringCharFeatures, BinaryLabels, DNA
from modshogun import WeightedDegreeStringKernel
try:
from modshogun import SVMLight
except ImportError:
print('No support for SVMLight available.')
return
feats_train = StringCharFeatures(DNA)
feats_train.set_features(fm_train_dna)
feats_test = StringCharFeatures(DNA)
feats_test.set_features(fm_test_dna)
degree = 20
kernel = WeightedDegreeStringKernel(feats_train, feats_train, degree)
labels = BinaryLabels(label_train_dna)
svm = SVMLight(C, kernel, labels)
svm.set_epsilon(epsilon)
svm.parallel.set_num_threads(num_threads)
svm.train()
kernel.init(feats_train, feats_test)
svm.apply().get_labels()
return kernel
def kernel_salzberg_word_string_modular(fm_train_dna=traindat,
fm_test_dna=testdat,
label_train_dna=label_traindat,
order=3,
gap=0,
reverse=False):
from modshogun import StringCharFeatures, StringWordFeatures, DNA, BinaryLabels
from modshogun import SalzbergWordStringKernel
from modshogun import PluginEstimate
charfeat = StringCharFeatures(fm_train_dna, DNA)
feats_train = StringWordFeatures(charfeat.get_alphabet())
feats_train.obtain_from_char(charfeat, order - 1, order, gap, reverse)
charfeat = StringCharFeatures(fm_test_dna, DNA)
feats_test = StringWordFeatures(charfeat.get_alphabet())
feats_test.obtain_from_char(charfeat, order - 1, order, gap, reverse)
pie = PluginEstimate()
labels = BinaryLabels(label_train_dna)
pie.set_labels(labels)
pie.set_features(feats_train)
pie.train()
kernel = SalzbergWordStringKernel(feats_train, feats_train, pie, labels)
km_train = kernel.get_kernel_matrix()
kernel.init(feats_train, feats_test)
pie.set_features(feats_test)
pie.apply().get_labels()
km_test = kernel.get_kernel_matrix()
return km_train, km_test, kernel
def runShogunSVMDNAWDNoPositionKernel(train_xt, train_lt, test_xt):
"""
run svm with non-position WD kernel
"""
##################################################
# set up svm
feats_train = StringCharFeatures(train_xt, DNA)
feats_test = StringCharFeatures(test_xt, DNA)
kernel = WeightedDegreeStringKernel(feats_train, feats_train, DEGREE)
kernel.io.set_loglevel(MSG_DEBUG)
weights=arange(1,DEGREE+1,dtype=double)[::-1]/ \
sum(arange(1,DEGREE+1,dtype=double))
kernel.set_wd_weights(weights)
# 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!"
out1 = svm.apply(feats_train).get_labels()
kernel.init(feats_train, feats_test)
out2 = svm.apply(feats_test).get_labels()
return out1, out2
def runShogunSVMDNALinearStringKernel(train_xt, train_lt, test_xt):
"""
run svm with spectrum kernel
"""
##################################################
# set up svm
feats_train = StringCharFeatures(train_xt, DNA)
feats_test = StringCharFeatures(test_xt, DNA)
kernel = LinearStringKernel(feats_train, feats_train)
kernel.io.set_loglevel(MSG_DEBUG)
# 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!"
out1 = svm.apply(feats_train).get_labels()
kernel.init(feats_train, feats_test)
out2 = svm.apply(feats_test).get_labels()
return out1, out2
def evaluation_cross_validation_classification(traindat=traindat,
label_traindat=label_traindat):
from modshogun import CrossValidation, CrossValidationResult
from modshogun import ContingencyTableEvaluation, ACCURACY
from modshogun import StratifiedCrossValidationSplitting
from modshogun import BinaryLabels
from modshogun import RealFeatures
from modshogun import LibLinear, L2R_L2LOSS_SVC
# training data
features = RealFeatures(traindat)
labels = BinaryLabels(label_traindat)
# classifier
classifier = LibLinear(L2R_L2LOSS_SVC)
# splitting strategy for 5 fold cross-validation (for classification its better
# to use "StratifiedCrossValidation", but the standard
# "CrossValidationSplitting" is also available
splitting_strategy = StratifiedCrossValidationSplitting(labels, 5)
# evaluation method
evaluation_criterium = ContingencyTableEvaluation(ACCURACY)
# cross-validation instance
cross_validation = CrossValidation(classifier, features, labels,
splitting_strategy,
evaluation_criterium)
cross_validation.set_autolock(False)
# (optional) repeat x-val 10 times
cross_validation.set_num_runs(10)
# perform cross-validation and print(results)
result = cross_validation.evaluate()
def classifier_svmocas_modular(train_fname=traindat,
test_fname=testdat,
label_fname=label_traindat,
C=0.9,
epsilon=1e-5,
num_threads=1):
from modshogun import RealFeatures, BinaryLabels
from modshogun import CSVFile
try:
from modshogun import SVMOcas
except ImportError:
print("SVMOcas not available")
return
feats_train = RealFeatures(CSVFile(train_fname))
feats_test = RealFeatures(CSVFile(test_fname))
labels = BinaryLabels(CSVFile(label_fname))
svm = SVMOcas(C, feats_train, labels)
svm.set_epsilon(epsilon)
svm.parallel.set_num_threads(num_threads)
svm.set_bias_enabled(False)
svm.train()
bias = svm.get_bias()
w = svm.get_w()
predictions = svm.apply(feats_test)
return predictions, svm, predictions.get_labels()
def kernel_histogram_word_string_modular (fm_train_dna=traindat,fm_test_dna=testdat,label_train_dna=label_traindat,order=3,ppseudo_count=1,npseudo_count=1): from modshogun import StringCharFeatures, StringWordFeatures, DNA, BinaryLabels from modshogun import HistogramWordStringKernel, AvgDiagKernelNormalizer from modshogun import PluginEstimate#, MSG_DEBUG charfeat=StringCharFeatures(DNA) #charfeat.io.set_loglevel(MSG_DEBUG) charfeat.set_features(fm_train_dna) feats_train=StringWordFeatures(charfeat.get_alphabet()) feats_train.obtain_from_char(charfeat, order-1, order, 0, False) charfeat=StringCharFeatures(DNA) charfeat.set_features(fm_test_dna) feats_test=StringWordFeatures(charfeat.get_alphabet()) feats_test.obtain_from_char(charfeat, order-1, order, 0, False) pie=PluginEstimate(ppseudo_count,npseudo_count) labels=BinaryLabels(label_train_dna) pie.set_labels(labels) pie.set_features(feats_train) pie.train() kernel=HistogramWordStringKernel(feats_train, feats_train, pie) km_train=kernel.get_kernel_matrix() kernel.init(feats_train, feats_test) pie.set_features(feats_test) pie.apply().get_labels() km_test=kernel.get_kernel_matrix() return km_train,km_test,kernel
def classifier_ssk_modular(fm_train_dna=traindat,
fm_test_dna=testdat,
label_train_dna=label_traindat,
C=1,
maxlen=1,
decay=1):
from modshogun import StringCharFeatures, BinaryLabels
from modshogun import LibSVM, StringSubsequenceKernel, DNA
from modshogun import ErrorRateMeasure
feats_train = StringCharFeatures(fm_train_dna, DNA)
feats_test = StringCharFeatures(fm_test_dna, DNA)
labels = BinaryLabels(label_train_dna)
kernel = StringSubsequenceKernel(feats_train, feats_train, maxlen, decay)
svm = LibSVM(C, kernel, labels)
svm.train()
out = svm.apply(feats_train)
evaluator = ErrorRateMeasure()
trainerr = evaluator.evaluate(out, labels)
# print(trainerr)
kernel.init(feats_train, feats_test)
predicted_labels = svm.apply(feats_test).get_labels()
# print predicted_labels
return predicted_labels
def modelselection_random_search_liblinear_modular (traindat=traindat, label_traindat=label_traindat):
from modshogun import CrossValidation, CrossValidationResult
from modshogun import ContingencyTableEvaluation, ACCURACY
from modshogun import StratifiedCrossValidationSplitting
from modshogun import RandomSearchModelSelection
from modshogun import ModelSelectionParameters, R_EXP
from modshogun import ParameterCombination
from modshogun import BinaryLabels
from modshogun import RealFeatures
from modshogun import LibLinear, L2R_L2LOSS_SVC
# build parameter tree to select C1 and C2
param_tree_root=ModelSelectionParameters()
c1=ModelSelectionParameters("C1");
param_tree_root.append_child(c1)
c1.build_values(-2.0, 2.0, R_EXP);
c2=ModelSelectionParameters("C2");
param_tree_root.append_child(c2);
c2.build_values(-2.0, 2.0, R_EXP);
# training data
features=RealFeatures(traindat)
labels=BinaryLabels(label_traindat)
# classifier
classifier=LibLinear(L2R_L2LOSS_SVC)
# print all parameter available for modelselection
# Dont worry if yours is not included but, write to the mailing list
#classifier.print_modsel_params()
# splitting strategy for cross-validation
splitting_strategy=StratifiedCrossValidationSplitting(labels, 10)
# evaluation method
evaluation_criterium=ContingencyTableEvaluation(ACCURACY)
# cross-validation instance
cross_validation=CrossValidation(classifier, features, labels,
splitting_strategy, evaluation_criterium)
cross_validation.set_autolock(False)
# model selection instance
model_selection=RandomSearchModelSelection(cross_validation, param_tree_root, 0.5)
# perform model selection with selected methods
#print "performing model selection of"
#param_tree_root.print_tree()
best_parameters=model_selection.select_model()
# print best parameters
#print "best parameters:"
#best_parameters.print_tree()
# apply them and print result
best_parameters.apply_to_machine(classifier)
result=cross_validation.evaluate()
def evaluation_cross_validation_mkl_weight_storage(
traindat=traindat, label_traindat=label_traindat):
from modshogun import CrossValidation, CrossValidationResult
from modshogun import CrossValidationPrintOutput
from modshogun import CrossValidationMKLStorage
from modshogun import ContingencyTableEvaluation, ACCURACY
from modshogun import StratifiedCrossValidationSplitting
from modshogun import BinaryLabels
from modshogun import RealFeatures, CombinedFeatures
from modshogun import GaussianKernel, CombinedKernel
from modshogun import LibSVM, MKLClassification
# training data, combined features all on same data
features = RealFeatures(traindat)
comb_features = CombinedFeatures()
comb_features.append_feature_obj(features)
comb_features.append_feature_obj(features)
comb_features.append_feature_obj(features)
labels = BinaryLabels(label_traindat)
# kernel, different Gaussians combined
kernel = CombinedKernel()
kernel.append_kernel(GaussianKernel(10, 0.1))
kernel.append_kernel(GaussianKernel(10, 1))
kernel.append_kernel(GaussianKernel(10, 2))
# create mkl using libsvm, due to a mem-bug, interleaved is not possible
svm = MKLClassification(LibSVM())
svm.set_interleaved_optimization_enabled(False)
svm.set_kernel(kernel)
# splitting strategy for 5 fold cross-validation (for classification its better
# to use "StratifiedCrossValidation", but the standard
# "StratifiedCrossValidationSplitting" is also available
splitting_strategy = StratifiedCrossValidationSplitting(labels, 5)
# evaluation method
evaluation_criterium = ContingencyTableEvaluation(ACCURACY)
# cross-validation instance
cross_validation = CrossValidation(svm, comb_features, labels,
splitting_strategy,
evaluation_criterium)
cross_validation.set_autolock(False)
# append cross vlaidation output classes
#cross_validation.add_cross_validation_output(CrossValidationPrintOutput())
mkl_storage = CrossValidationMKLStorage()
cross_validation.add_cross_validation_output(mkl_storage)
cross_validation.set_num_runs(3)
# perform cross-validation
result = cross_validation.evaluate()
# print mkl weights
weights = mkl_storage.get_mkl_weights()
def load_sparse_data(filename, dimension=None):
input_file = LibSVMFile(args.dataset)
sparse_feats = SparseRealFeatures()
label_array = sparse_feats.load_with_labels(input_file)
labels = BinaryLabels(label_array)
if dimension != None:
sparse_feats.set_num_features(dimension)
return {'data': sparse_feats, 'labels': labels}
def evaluation_contingencytableevaluation_modular(ground_truth, predicted):
from modshogun import BinaryLabels
from modshogun import ContingencyTableEvaluation
from modshogun import AccuracyMeasure, ErrorRateMeasure, BALMeasure
from modshogun import WRACCMeasure, F1Measure, CrossCorrelationMeasure
from modshogun import RecallMeasure, PrecisionMeasure, SpecificityMeasure
ground_truth_labels = BinaryLabels(ground_truth)
predicted_labels = BinaryLabels(predicted)
base_evaluator = ContingencyTableEvaluation()
base_evaluator.evaluate(predicted_labels, ground_truth_labels)
evaluator = AccuracyMeasure()
accuracy = evaluator.evaluate(predicted_labels, ground_truth_labels)
evaluator = ErrorRateMeasure()
errorrate = evaluator.evaluate(predicted_labels, ground_truth_labels)
evaluator = BALMeasure()
bal = evaluator.evaluate(predicted_labels, ground_truth_labels)
evaluator = WRACCMeasure()
wracc = evaluator.evaluate(predicted_labels, ground_truth_labels)
evaluator = F1Measure()
f1 = evaluator.evaluate(predicted_labels, ground_truth_labels)
evaluator = CrossCorrelationMeasure()
crosscorrelation = evaluator.evaluate(predicted_labels,
ground_truth_labels)
evaluator = RecallMeasure()
recall = evaluator.evaluate(predicted_labels, ground_truth_labels)
evaluator = PrecisionMeasure()
precision = evaluator.evaluate(predicted_labels, ground_truth_labels)
evaluator = SpecificityMeasure()
specificity = evaluator.evaluate(predicted_labels, ground_truth_labels)
return accuracy, errorrate, bal, wracc, f1, crosscorrelation, recall, precision, specificity
def runShogunSVMDNACombinedSpectrumKernel(train_xt, train_lt, test_xt):
"""
run svm with combined spectrum kernel
"""
##################################################
# set up svm
kernel=CombinedKernel()
feats_train=CombinedFeatures()
feats_test=CombinedFeatures()
for K in KList:
# Iterate through the K's and make a spectrum kernel for each
charfeat_train = StringCharFeatures(train_xt, DNA)
current_feats_train = StringWordFeatures(DNA)
current_feats_train.obtain_from_char(charfeat_train, K-1, K, GAP, False)
preproc=SortWordString()
preproc.init(current_feats_train)
current_feats_train.add_preprocessor(preproc)
current_feats_train.apply_preprocessor()
feats_train.append_feature_obj(current_feats_train)
charfeat_test = StringCharFeatures(test_xt, DNA)
current_feats_test=StringWordFeatures(DNA)
current_feats_test.obtain_from_char(charfeat_test, K-1, K, GAP, False)
current_feats_test.add_preprocessor(preproc)
current_feats_test.apply_preprocessor()
feats_test.append_feature_obj(current_feats_test)
current_kernel=CommWordStringKernel(10, False)
kernel.append_kernel(current_kernel)
kernel.io.set_loglevel(MSG_DEBUG)
# init kernel
labels = BinaryLabels(train_lt)
# run svm model
print "Ready to train!"
kernel.init(feats_train, feats_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 get_labels(raw=False, type='binary'):
data = concatenate(
array(
(-ones(NUM_EXAMPLES, dtype=double), ones(NUM_EXAMPLES,
dtype=double))))
if raw:
return data
else:
if type == 'binary':
return BinaryLabels(data)
if type == 'regression':
return RegressionLabels(data)
return None
def kernel_auc_modular(train_fname=traindat,
label_fname=label_traindat,
width=1.7):
from modshogun import GaussianKernel, AUCKernel, RealFeatures
from modshogun import BinaryLabels, CSVFile
feats_train = RealFeatures(CSVFile(train_fname))
subkernel = GaussianKernel(feats_train, feats_train, width)
kernel = AUCKernel(0, subkernel)
kernel.setup_auc_maximization(BinaryLabels(CSVFile(label_fname)))
km_train = kernel.get_kernel_matrix()
return kernel
def runShogunOneClassSVMDNASpectrumKernel(train_xt, train_lt, test_xt):
"""
run svm with spectrum kernel
"""
##################################################
# set up svr
charfeat_train = StringCharFeatures(train_xt, DNA)
feats_train = StringWordFeatures(DNA)
feats_train.obtain_from_char(charfeat_train, K-1, K, GAP, False)
preproc=SortWordString()
preproc.init(feats_train)
feats_train.add_preprocessor(preproc)
feats_train.apply_preprocessor()
charfeat_test = StringCharFeatures(test_xt, DNA)
feats_test=StringWordFeatures(DNA)
feats_test.obtain_from_char(charfeat_test, K-1, K, GAP, False)
feats_test.add_preprocessor(preproc)
feats_test.apply_preprocessor()
kernel=CommWordStringKernel(feats_train, feats_train, False)
kernel.io.set_loglevel(MSG_DEBUG)
# init kernel
labels = BinaryLabels(train_lt)
# run svm model
print "Ready to train!"
svm=LibSVMOneClass(SVMC, kernel)
svm.set_epsilon(EPSILON)
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()
# predictions = svm.apply(feats_test)
# return predictions, svm, predictions.get_labels()
return out1,out2,out1DecisionValues,out2DecisionValues
#!/usr/bin/env python2.7 # # This software is distributed under BSD 3-clause license (see LICENSE file). # # Copyright (C) 2014 Thoralf Klein # from modshogun import RealFeatures, BinaryLabels, LibLinear from numpy import random, mean X_train = RealFeatures(random.randn(30, 100)) Y_train = BinaryLabels(random.randn(X_train.get_num_vectors())) svm = LibLinear(1.0, X_train, Y_train) svm.train() Y_pred = svm.apply_binary(X_train) Y_train.get_labels() == Y_pred.get_labels() print "accuracy:", mean(Y_train.get_labels() == Y_pred.get_labels())
#!/usr/bin/env python2.7
#
# This software is distributed under BSD 3-clause license (see LICENSE file).
#
# Copyright (C) 2014 Thoralf Klein
#
from modshogun import RealFeatures, BinaryLabels
from modshogun import LibLinear, L2R_L2LOSS_SVC_DUAL
from numpy import random, mean
X_train = RealFeatures(random.randn(30, 100))
Y_train = BinaryLabels(random.randn(X_train.get_num_vectors()))
results = []
for C1_pow in range(-3, 1):
for C2_pow in range(-3, 1):
svm = LibLinear()
svm.set_bias_enabled(False)
svm.set_liblinear_solver_type(L2R_L2LOSS_SVC_DUAL)
svm.set_C(10**C1_pow, 10**C2_pow)
svm.set_features(X_train)
svm.set_labels(Y_train)
svm.train()
Y_pred = svm.apply_binary(X_train)
accuracy = mean(Y_train.get_labels() == Y_pred.get_labels())