def __assign_shogun_version():
"""Assign shogun versions
"""
if 'shogun' in versions:
return
import shogun.Classifier as __sc
versions['shogun:rev'] = __sc.Version_get_version_revision()
ver = __sc.Version_get_version_release().lstrip('v')
versions['shogun:full'] = ver
if '_' in ver:
ver = ver[:ver.index('_')]
versions['shogun'] = ver
def _run_lda ():
"""Run Linear Discriminant Analysis classifier."""
params={
'name': 'LDA',
'type': 'lda',
'gamma': 0.1,
'num_threads': 1,
'data': dataop.get_clouds(2),
'feature_class': 'simple',
'feature_type': 'Real',
'label_type': 'twoclass',
'accuracy': 1e-7
}
feats=featop.get_features(
params['feature_class'], params['feature_type'], params['data'])
params['labels'], labels=dataop.get_labels(
feats['train'].get_num_vectors(), params['label_type'])
lda=classifier.LDA(params['gamma'], feats['train'], labels)
lda.parallel.set_num_threads(params['num_threads'])
lda.train()
lda.set_features(feats['test'])
params['classified']=lda.classify().get_labels()
output=fileop.get_output(category.CLASSIFIER, params)
fileop.write(category.CLASSIFIER, output)
def _run_perceptron ():
"""Run Perceptron classifier."""
params={
'name': 'Perceptron',
'type': 'perceptron',
'num_threads': 1,
'learn_rate': .1,
'max_iter': 1000,
'data': dataop.get_clouds(2),
'feature_class': 'simple',
'feature_type': 'Real',
'label_type': 'twoclass',
'accuracy': 1e-7
}
feats=featop.get_features(
params['feature_class'], params['feature_type'], params['data'])
num_vec=feats['train'].get_num_vectors()
params['labels'], labels=dataop.get_labels(num_vec, params['label_type'])
perceptron=classifier.Perceptron(feats['train'], labels)
perceptron.parallel.set_num_threads(params['num_threads'])
perceptron.set_learn_rate(params['learn_rate'])
perceptron.set_max_iter(params['max_iter'])
perceptron.train()
params['bias']=perceptron.get_bias()
perceptron.set_features(feats['test'])
params['classified']=perceptron.classify().get_labels()
output=fileop.get_output(category.CLASSIFIER, params)
fileop.write(category.CLASSIFIER, output)
def bench_shogun(X, y, T, valid):
#
# .. Shogun ..
#
from shogun import Classifier, Features, Distance
start = datetime.now()
feat = Features.RealFeatures(X.T)
distance = Distance.EuclidianDistance(feat, feat)
labels = Features.Labels(y.astype(np.float64))
test_feat = Features.RealFeatures(T.T)
knn = Classifier.KNN(n_neighbors, distance, labels)
knn.train()
score = np.mean(knn.classify(test_feat).get_labels() == valid)
return score, datetime.now() - start
def __check_shogun(bottom_version, custom_versions=[]):
"""Check if version of shogun is high enough (or custom known) to
be enabled in the testsuite
Parameters
----------
bottom_version : int
Bottom version which must be satisfied
custom_versions : list of int
Arbitrary list of versions which could got patched for
a specific issue
"""
import shogun.Classifier as __sc
ver = __sc.Version_get_version_revision()
__assign_shogun_version()
if (ver in custom_versions) or (ver >= bottom_version):
return True
else:
raise ImportError, 'Version %s is smaller than needed %s' % \
(ver, bottom_version)
def _run_knn ():
"""Run K-Nearest-Neighbour classifier.
"""
params={
'name': 'EuclidianDistance',
'data': dataop.get_clouds(2),
'feature_class': 'simple',
'feature_type': 'Real'
}
feats=featop.get_features(
params['feature_class'], params['feature_type'], params['data'])
dfun=eval(params['name'])
distance=dfun(feats['train'], feats['train'])
output=fileop.get_output(category.DISTANCE, params)
params={
'name': 'KNN',
'type': 'knn',
'num_threads': 1,
'k': 3,
'label_type': 'twoclass',
'accuracy': 1e-8
}
params['labels'], labels=dataop.get_labels(
feats['train'].get_num_vectors(), params['label_type'])
knn=classifier.KNN(params['k'], distance, labels)
knn.parallel.set_num_threads(params['num_threads'])
knn.train()
distance.init(feats['train'], feats['test'])
params['classified']=knn.classify().get_labels()
output.update(fileop.get_output(category.CLASSIFIER, params))
fileop.write(category.CLASSIFIER, output)
def parse_config_file(config_files, feature_file_suffix='.bed'):
""" Return classifiers. """
valid_words = ['CLASSIFIER', 'KERNEL', 'KERNEL_NAME', 'KERNEL_NORM', 'FEATURE',
'BIN_FEATURE', 'CON_FEATURE', 'PP_FEATURE', 'SEQ', 'END', 'C', 'MKL_NORM', 'REV_COMP']
classifiers = []
for config_file in config_files:
name = None; kernels = []; c = None
bin_features = []; con_features = []; pp_features = []
seqs = []; ks = [];
kern_names = []
kern_norms = []
rev_comps = []
mkl_norm = 2
line_num = 0
for line in open(config_file):
line_num += 1
line = line.strip()
if line.startswith('#') or not line: continue
t = line.split()
if t[0] not in valid_words:
sys.exit('ERROR! %s not recognized in line %d: %s.' % (t[0], line_num, line))
elif t[0] == 'CLASSIFIER':
name = '_'.join(t[1:])
elif t[0] == 'KERNEL':
bin_features.append([])
con_features.append([])
pp_features.append([])
seqs.append([])
kern_norms.append(None)
kern_names.append(None)
rev_comps.append(False)
ks.append(None)
if t[1] == 'Linear':
kernels.append(LinearKernel())
elif t[1].startswith('CommWordString_') or t[1].startswith('Spectrum_'):
k = int(t[1].split('_')[-1])
ks[-1] = k
kernels.append(CommWordStringKernel(10, False))
elif t[1].startswith('Gaussian_'):
sigma = float(t[1].split('_')[-1])
kernels.append(GaussianKernel(10, sigma))
else:
sys.exit('ERROR! %s is not a valid kernel.' % t[1])
elif t[0] == 'KERNEL_NORM':
if t[1] == 'VarianceKernelNormalizer':
kern_norms[-1] = VarianceKernelNormalizer()
kernels[-1].set_normalizer(VarianceKernelNormalizer())
elif t[1] == 'SqrtDiagKernelNormalizer':
kern_norms[-1] = SqrtDiagKernelNormalizer()
kernels[-1].set_normalizer(SqrtDiagKernelNormalizer())
elif t[1] == 'AvgDiagKernelNormalizer':
kern_norms[-1] = AvgDiagKernelNormalizer()
kernels[-1].set_normalizer(AvgDiagKernelNormalizer())
else:
sys.exit('ERROR! %s is not a recognized kernel normalizer.' % t[1])
elif t[0] == 'KERNEL_NAME':
kern_names[-1] = t[1]
elif t[0] == 'REV_COMP':
rev_comps[-1] = bool(int(t[1]))
elif t[0] == 'BIN_FEATURE' or t[0] == 'FEATURE': # for BW compatibility
bf_path = t[1]
if os.path.isdir(bf_path):
for fn in os.listdir(bf_path):
if not fn.endswith(feature_file_suffix): continue
full_fn = bf_path + ('/' if bf_path[-1] != '/' else '') + fn
if full_fn not in bin_features[-1]: bin_features[-1].append(full_fn)
else:
bin_features[-1].append(bf_path)
elif t[0] == 'CON_FEATURE':
cf_path = t[1]
if os.path.isdir(cf_path):
for fn in os.listdir(cf_path):
if not fn.endswith(feature_file_suffix): continue
full_fn = cf_path + ('/' if cf_path[-1] != '/' else '') + fn
if full_fn not in con_features[-1]: con_features[-1].append(full_fn)
else:
con_features[-1].append(cf_path)
elif t[0] == 'PP_FEATURE':
pf_path = t[1]
if os.path.isdir(pf_path):
for fn in os.listdir(pf_path):
if not fn.endswith(feature_file_suffix): continue
full_fn = pf_path + ('/' if pf_path[-1] != '/' else '') + fn
if full_fn not in pp_features[-1]: pp_features[-1].append(full_fn)
else:
pp_features[-1].append(pf_path)
elif t[0] == 'SEQ':
seqs[-1].append(t[1])
elif t[0] == 'C':
c = [float(x) for x in t[1:]]
elif t[0] == 'MKL_NORM':
mkl_norm = int(t[1])
elif t[0] == 'END':
print '\t', name
classifiers.append(Classifier(name, kernels, bin_features, con_features,
pp_features=pp_features, seqs=seqs,
kern_norms=kern_norms, ks=ks, c=c,
kern_names=kern_names,
mkl_norm=mkl_norm, rev_comps=rev_comps))
name = None; kernels = []; c = None
bin_features = []; con_features; pp_features = []
seqs = []; ks = []
kern_norms = []
kern_names = []
rev_comps = []
mkl_norm = 2
return classifiers
def train_attribute(attribute_id, C, split=0):
from shogun import Classifier, Features, Kernel, Distance
attribute_id = int(attribute_id)
print "# attribute ", attributenames[attribute_id]
C = float(C)
print "# C ", C
if split == 0:
train_classes = loadstr(
'/nfs3group/chlgrp/datasets/Animals_with_Attributes/trainclasses.txt'
)
test_classes = loadstr(
'/nfs3group/chlgrp/datasets/Animals_with_Attributes/testclasses.txt'
)
else:
classnames = loadstr(
'/nfs3group/chlgrp/datasets/Animals_with_Attributes/classnames.txt'
)
startid = (split - 1) * 10
stopid = split * 10
test_classes = classnames[startid:stopid]
train_classes = classnames[0:startid] + classnames[stopid:]
Xtrn, Ltrn = create_data(train_classes, attribute_id)
Xtst, Ltst = create_data(test_classes, attribute_id)
if min(Ltrn) == max(Ltrn): # only 1 class
Lprior = mean(Ltrn)
prediction = sign(Lprior) * ones(len(Ltst))
probabilities = 0.1 + 0.8 * 0.5 * (Lprior + 1.) * ones(
len(Ltst)) # fallback
return prediction, probabilities, Ltst
#sg('loglevel', 'WARN')
widths = {}
for feature in all_features:
traindata = array(Xtrn[feature][:, ::50], float) # used to be 5*offset
trainfeat = Features.RealFeatures(traindata)
DM = Distance.ChiSquareDistance(trainfeat,
trainfeat).get_distance_matrix()
widths[feature] = median(DM.flatten())
del traindata, trainfeat, DM
s = Classifier.LibSVM() #sg('new_svm', 'LIBSVM')
#sg('use_mkl', False) # we use fixed weights here
#sg('clean_features', 'TRAIN')
#sg('clean_features', 'TEST')
Lplatt_trn = concatenate([Ltrn[i::10]
for i in range(9)]) # 90% for training
Lplatt_val = Ltrn[9::10] # remaining 10% for platt scaling
feats_trn = Features.CombinedFeatures()
feats_val = Features.CombinedFeatures()
for feature in all_features:
Xplatt_trn = concatenate([Xtrn[feature][:, i::10] for i in range(9)],
axis=1)
feats_trn.append_feature_obj(
Features.RealFeatures(ascontiguousarray(Xplatt_trn)))
#sg('add_features', 'TRAIN', Xplatt_trn)
Xplatt_val = Xtrn[feature][:, 9::10]
feats_val.append_feature_obj(
Features.RealFeatures(ascontiguousarray(Xplatt_val)))
#sg('add_features', 'TEST', Xplatt_val)
del Xplatt_trn, Xplatt_val, Xtrn[feature]
labels_trn = Features.Labels(Lplatt_trn)
#sg('set_labels', 'TRAIN', Lplatt_trn)
kernel = Kernel.CombinedKernel()
#sg('set_kernel', 'COMBINED', 5000)
for featureset in all_features:
kernel.append_kernel(Kernel.Chi2Kernel(5000, widths[featureset] / 5.))
#sg('add_kernel', 1., 'CHI2', 'REAL', 10, widths[featureset]/5. )
kernel.init(feats_trn, feats_trn)
K = kernel.get_kernel_matrix()
K.tofile('/scratch/chl/cvfold%d_C%g_%02d-trn.kernel' %
(split, C, attribute_id))
del K
s.set_max_train_time(600 * 60.)
#sg('svm_max_train_time', 600*60.) # one hour should be plenty
s.set_C(C, C)
#sg('c', C)
s.set_kernel(kernel)
s.set_labels(labels_trn)
#sg('init_kernel', 'TRAIN')
try:
s.train()
#sg('train_classifier')
except (RuntimeWarning, RuntimeError
): # can't train, e.g. all samples have the same labels
Lprior = mean(Ltrn)
prediction = sign(Lprior) * ones(len(Ltst))
probabilities = 0.1 + 0.8 * 0.5 * (Lprior + 1.) * ones(len(Ltst))
savetxt('./DAP/cvfold%d_C%g_%02d.txt' % (split, C, attribute_id),
prediction)
savetxt('./DAP/cvfold%d_C%g_%02d.prob' % (split, C, attribute_id),
probabilities)
savetxt('./DAP/cvfold%d_C%g_%02d.labels' % (split, C, attribute_id),
Ltst)
return prediction, probabilities, Ltst
bias = s.get_bias()
alphas = s.get_alphas()
#[bias, alphas]=sg('get_svm')
#print bias,alphas
kernel.init(feats_trn, feats_val)
K = kernel.get_kernel_matrix()
K.tofile('/scratch/chl/cvfold%d_C%g_%02d-val.kernel' %
(split, C, attribute_id))
del K
#sg('init_kernel', 'TEST')
try:
prediction = s.classify().get_labels()
#prediction=sg('classify')
platt_params = SigmoidTrain(prediction, Lplatt_val)
probabilities = SigmoidPredict(prediction, platt_params)
savetxt('./DAP/cvfold%d_C%g_%02d-val.txt' % (split, C, attribute_id),
prediction)
savetxt('./DAP/cvfold%d_C%g_%02d-val.prob' % (split, C, attribute_id),
probabilities)
savetxt(
'./DAP/cvfold%d_C%g_%02d-val.labels' % (split, C, attribute_id),
Lplatt_val)
savetxt('./DAP/cvfold%d_C%g_%02d-val.platt' % (split, C, attribute_id),
platt_params)
#print '#train-perf ',attribute_id,C,mean((prediction*Lplatt_val)>0),mean(Lplatt_val>0)
#print '#platt-perf ',attribute_id,C,mean((sign(probabilities-0.5)*Lplatt_val)>0),mean(Lplatt_val>0)
except RuntimeError:
Lprior = mean(Ltrn)
prediction = sign(Lprior) * ones(len(Ltst))
probabilities = 0.1 + 0.8 * 0.5 * (Lprior + 1.) * ones(len(Ltst))
print >> sys.stderr, "#Error during testing. Using constant platt scaling"
platt_params = [1., 0.]
# ----------------------------- now apply to test classes ------------------
feats_tst = Features.CombinedFeatures()
#sg('clean_features', 'TEST')
for feature in all_features:
feats_tst.append_feature_obj(
Features.RealFeatures(ascontiguousarray(Xtst[feature])))
del Xtst[feature]
kernel.init(feats_trn, feats_tst)
K = kernel.get_kernel_matrix()
K.tofile('/scratch/chl/cvfold%d_C%g_%02d-tst.kernel' %
(split, C, attribute_id))
del K
#sg('init_kernel', 'TEST')
prediction = s.classify().get_labels()
#prediction=sg('classify')
probabilities = SigmoidPredict(prediction, platt_params)
savetxt('./DAP/cvfold%d_C%g_%02d.txt' % (split, C, attribute_id),
prediction)
savetxt('./DAP/cvfold%d_C%g_%02d.prob' % (split, C, attribute_id),
probabilities)
savetxt('./DAP/cvfold%d_C%g_%02d.labels' % (split, C, attribute_id), Ltst)
#print '#test-perf ',attribute_id,C,mean((prediction*Ltst)>0),mean(Ltst>0)
#print '#platt-perf ',attribute_id,C,mean((sign(probabilities-0.5)*Ltst)>0),mean(Ltst>0)
return prediction, probabilities, Ltst