def __init__(self, task='spam'):
super(TaskTrainer, self).__init__()
from sklearn.svm import SVC as SVM
self.task = task
if task == 'vehicle':
self.env = SVM(C=1e2, kernel='rbf', random_state=0) # For vehicle task
elif task == 'page':
self.env = SVM(C=1e2, kernel='rbf', random_state=0, gamma=1e-2) # For page blocks
elif task == 'credit':
self.env = DT(max_depth=4) # For credit card task
elif task == 'spam':
self.env = LogisticRegression(C=1e2, random_state=0) # For spam detection task
def __init__(self, train_X, test_X, train_Y, test_Y, agent, classifier,
save_conf_mat):
self.train_X = train_X
self.test_X = test_X
self.train_Y = train_Y
self.test_Y = test_Y
self.classifier = classifier
if (self.classifier.lower() == 'knn'):
self.clf = KNN()
elif (self.classifier.lower() == 'rf'):
self.clf = RF()
elif (self.classifier.lower() == 'svm'):
self.clf = SVM()
else:
self.clf = None
print('\n[Error!] We don\'t currently support {} classifier...\n'.
format(classifier))
exit(1)
if (agent == None):
self.agent = np.ones(train_X.shape[1])
self.predictions = self.classify()
self.accuracy = self.compute_accuracy()
self.precision = self.compute_precision()
self.recall = self.compute_recall()
self.f1_score = self.compute_f1()
self.confusion_matrix = self.compute_confusion_matrix()
self.plot_confusion_matrix(save_conf_mat)
def main(x, y, task):
#ys = [yr, ym, y25]
#y_names = ['readm', 'mort_h', 'pheno25']
#xs = [x48, onehot, w2v, w48, sentences]
#x_names = ['48h', 'sparse_dx', 'w2v', 'w2v_48h', 'sentences']
lr = LR(C=1e-4, penalty='l2', verbose=1) #sag if multiclass/multilabel
svm = SVM(C=1e5, verbose=True)
rf = RF(n_estimators=60, verbose=1)
gbc = GBC(n_estimators=200, learning_rate=1e-3, verbose=1)
models = [lr, svm, rf, gbc]
names = ['LR', 'SVM', 'RF', 'GBC']
data = {}
for idx in range(len(models)):
if task != 'binary':
data[names[idx]] = {}
for ix in range(25):
dat = run_experiment(x, y[:, ix], models[idx], task)
data[names[idx]][ix] = dat
else:
dat = run_experiment(x, y, models[idx], task)
data[names[idx]] = dat
return (data)
def _build_target_classifier(args, samples, one_vs_all_labels,
transferability_values):
compute_kernel = lambda x, y: _compute_target_classifier_kernel_matrix(
args, samples, one_vs_all_labels, transferability_values)
classifier = SVM(kernel=compute_kernel)
classifier.fit(samples, one_vs_all_labels)
return classifier
def __init__(self,
kernel='rbf',
gamma='scale',
tol=0.001,
nu=0.5,
shrinking=True,
max_iter=1000):
"""
Unsupervised Outlier Detection.
Arguments
---------
kernel : {‘linear’, ‘poly’, ‘rbf’, ‘sigmoid’, ‘precomputed’}, optional (default=rbf).
Specifies the kernel type to be used in the algorithm.
It must be one of ‘linear’, ‘poly’, ‘rbf’, ‘sigmoid’, ‘precomputed’ or a callable
gamma : {‘scale’, ‘auto’} or float, default=’scale’
Kernel coefficient for ‘rbf’, ‘poly’ and ‘sigmoid’.
tol : float, default=1e-3
Tolerance for stopping criterion
nu : float, default=0.5
An upper bound on the fraction of training errors and a lower bound of the fraction of support vectors.
Should be in the interval (0, 1]. By default 0.5 will be taken
max_iter : int, default=-1
Hard limit on iterations within solver, or -1 for no limit.
Reference
---------
For more information, please visit https://scikit-learn.org/stable/modules/generated/sklearn.svm.OneClassSVM.html
"""
self.model = SVM(kernel=kernel,
gamma=gamma,
tol=tol,
nu=nu,
shrinking=shrinking,
max_iter=max_iter)
self.transformer = None
def train_model(param):
model= SVM()
C = 10**(param)
print("Param C= ",C)
model.set_params(C=C,kernel='linear')
model.fit(X_train,Y_train)
print("Test Accuracy: ",model.score(X_test,Y_test))
return model
def fit(self, X, y):
for C in self.c_values:
svm = SVM(dual=False, penalty="l1", C=C, class_weight='auto')
svm.fit(X, y)
if np.sum(svm.coef_!=0) >= self.n_non_null:
self.svm = svm
self.coef_ = self.svm.coef_
self.C = C
break
return self
def optimal_svm_kernel (X_train, y_train, X_validate, y_validate):
best_kernel = None
best_accuracy = 0
for kernel in ['linear', 'rbf', 'poly', 'sigmoid']:
classifier = SVM(X_train, y_train, kernel)
accuracy = classifier.score(X_validate, y_validate)
if best_kernel is None or accuracy > best_accuracy:
best_accuracy = accuracy
best_kernel = kernel
print kernel, ":", accuracy
return best_kernel, best_accuracy
def _get_source_class_classifiers(args, all_samples):
source_class_classifiers = [
] # To find a mapping between classifier and label, just use source_class_classifiers[i] and source_class_list[i]
for source_class in args.source_class_list:
one_vs_all_labels = [0] * len(all_samples)
for i in range(len(args.source_samples)):
one_vs_all_labels[
i] = 1 if args.source_labels[i] == source_class else 0
classifier = SVM(kernel="linear")
classifier.fit(all_samples, one_vs_all_labels)
source_class_classifiers.append(classifier)
return source_class_classifiers
def train(self):
logging.info('-' * 20)
logging.info('Start training the %s model', self.model)
train_data = self.feature_extractor.extract_feature(
self.data_loader.get_trainset())
if self.model == 'GNB':
# Gaussian naive bayes
self.classifier = GNB()
elif self.model == 'BNB':
# Bernoulli naive bayes
self.classifier = BNB()
# self.tok = RT(r'\w+')
# vectorizer = Vectorizer(tokenizer=self.tok.tokenize)
# train_data = self.data_loader.get_trainset()
# train_data = [vectorizer.fit_transform(train_data[0]).toarray(), train_data[1]]
# self.vocabulary = vectorizer.get_feature_names()
elif self.model == 'MNB':
# Multinomial naive bayes
self.classifier = MNB()
elif self.model == 'LR':
# Logistic regression
param = {'C': [10, 5, 2, 1, 0.5, 0.2, 0.1, 0.05, 0.02, 0.01]}
self.classifier = GS(cv=5,
estimator=LR(penalty=self.penalty,
max_iter=self.epoch,
solver='liblinear'),
param_grid=param)
elif self.model == 'SVM':
# Support vector machine
self.penalty = self.penalty if self.penalty in ['l1', 'l2'
] else 'l2'
dual = self.penalty == 'l2'
#self.classifier = SVM(penalty=self.penalty, C=self.c, max_iter=self.epoch, dual=dual)
param = {'C': [10, 5, 2, 1, 0.5, 0.2, 0.1, 0.05, 0.02, 0.01]}
self.classifier = GS(cv=5,
estimator=SVM(penalty=self.penalty,
dual=dual,
max_iter=self.epoch),
param_grid=param)
elif self.model == 'R':
# RandomGuess
self.classifier = DC(strategy='stratified')
else:
logging.info('Unsupported model : %s', self.model)
exit(0)
self.classifier.fit(train_data[0], train_data[1])
self.classifier.predict(train_data[0])
predictions = self.classifier.predict(train_data[0])
acc = evaluator.accuracy_score(train_data[1], predictions)
return acc
def iterate(cself, svm, selectors, instances, K):
cself.mention('Training SVM...')
alphas, obj = qp.solve(cself.verbose)
# Construct SVM from solution
svm = SVM(kernel=self.kernel,
gamma=self.gamma,
p=self.p,
verbose=self.verbose,
sv_cutoff=self.sv_cutoff)
svm._X = instances
svm._y = classes
svm._alphas = alphas
svm._objective = obj
svm._compute_separator(K)
svm._K = K
cself.mention('Recomputing classes...')
p_confs = svm.predict(bs.pos_instances)
pos_selectors = bs.L_n + np.array([
l + np.argmax(p_confs[l:u])
for l, u in slices(bs.pos_groups)
])
new_selectors = np.hstack([neg_selectors, pos_selectors])
if selectors is None:
sel_diff = len(new_selectors)
else:
sel_diff = np.nonzero(new_selectors -
selectors)[0].size
cself.mention('Selector differences: %d' % sel_diff)
if sel_diff == 0:
return None, svm
elif sel_diff > 5:
# Clear results to avoid a
# bad starting point in
# the next iteration
qp.clear_results()
cself.mention('Updating QP...')
indices = (new_selectors, )
K = K_all[indices].T[indices].T
D = spdiag(classes)
qp.update_H(D * K * D)
return {
'svm': svm,
'selectors': new_selectors,
'instances': bs.instances[indices],
'K': K
}, None
def __init__(self, train_X, test_X, train_Y, test_Y, agent, classifier,
save_conf_mat, averaging):
self.agent = agent
if self.agent is None:
self.agent = np.ones(train_X.shape[1])
cols = np.flatnonzero(self.agent)
if (cols.shape[0] == 0):
print('[Error!] There are 0 features in the agent......')
exit(1)
# store the train and test features and labels
self.train_X = train_X[:, cols]
self.test_X = test_X[:, cols]
self.train_Y = train_Y
self.test_Y = test_Y
# set the classifier type
self.classifier = classifier
# get the unique labels
self.labels = np.unique(train_Y)
# select the averaging procedure
if (len(self.labels) == 2):
self.averaging = "binary"
else:
self.averaging = averaging
# setup the classifier
if (self.classifier.lower() == 'knn'):
self.clf = KNN()
elif (self.classifier.lower() == 'rf'):
self.clf = RF()
elif (self.classifier.lower() == 'svm'):
self.clf = SVM()
else:
self.clf = None
print('\n[Error!] We don\'t currently support {} classifier...\n'.
format(classifier))
exit(1)
# call the member functions
self.predictions = self.classify()
self.accuracy = self.compute_accuracy()
self.precision = self.compute_precision()
self.recall = self.compute_recall()
self.f1_score = self.compute_f1()
self.confusion_matrix = self.compute_confusion_matrix()
self.plot_confusion_matrix(save_conf_mat)
def make_model(opts, input_shape, aux_shape):
targets, multiclass, deep = get_setup(opts)
if opts.model == 'lstm':
if aux_shape:
model = hierarchical_lstm(input_shape=input_shape, aux_shape=aux_shape,
targets=targets, hidden = opts.hidden_size,
multiclass= multiclass, learn_rate=opts.learning_rate)
else:
model = lstm_model(input_shape=input_shape, targets=targets, hidden = opts.hidden_size,
multiclass= multiclass, learn_rate=opts.learning_rate)
elif opts.model == 'cnn':
if aux_shape:
model = hierarchical_cnn(input_shape=input_shape, aux_shape=aux_shape,
targets=targets, hidden = opts.hidden_size,
multiclass= multiclass, learn_rate=opts.learning_rate)
else:
model = cnn_model(input_shape=input_shape, targets=targets, hidden = opts.hidden_size,
multiclass= multiclass, learn_rate=opts.learning_rate)
elif opts.model == 'mlp':
model = mlp_model(input_shape=input_shape, targets=targets, hidden = opts.hidden_size,
multiclass= multiclass, learn_rate=opts.learning_rate)
elif opts.model == 'svm':
if targets>1: model = OneVsRestClassifier(SVM(C=50000, kernel = 'rbf', max_iter= 1000, verbose = True, decision_function_shape = 'ovr', probability = True))
else: model = SVM(C=1e4, kernel = 'linear', verbose = True, probability = True, max_iter= 1000)
elif opts.model == 'rf':
if targets>1: model = RF(n_estimators = 450, class_weight = 'balanced', criterion = 'entropy', bootstrap = False, verbose = 1)
else: model = RF(n_estimators = 450, verbose = 1)
elif opts.model =='gbc':
if targets>1: model = OneVsRestClassifier(GBC(n_estimators = 484, learning_rate = 0.0984, verbose = 1))
else: model = GBC(n_estimators = 400, learning_rate = 0.09, verbose = 1)
elif opts.model == 'lr':
if targets>1: model = OneVsRestClassifier(LR(max_iter= 1000, class_weight = 'balanced', multi_class = 'ovr', C = .09, penalty = 'l1', verbose = 1)) #sag if multiclass/multilabel
else: model = LR(C = 1e-3, penalty = 'l2', verbose = 1) #sag if multiclass/multilabel
else:
model = None
return model
def classifySVM(Xtrain, Ytrain, Xval, Yval):
# set range of C parameters to test
#cvals = np.logspace(-2,7,40)
cvals = [405]
# set range of gamma values to test
#gammas = np.logspace(-10,2,40)
gammas = [0.00015]
# find most accurate combination of C, gamma
best_accuracy = 0
best_model = None
best_c = None
best_g = None
# cycle through all combinations of C, gamma
for c in cvals:
for g in gammas:
clf = SVM(C=c,
cache_size=200,
class_weight=None,
coef0=0.0,
degree=3,
gamma=g,
kernel='rbf',
max_iter=-1,
probability=False,
random_state=None,
shrinking=True,
tol=0.001,
verbose=False)
clf.fit(Xtrain, Ytrain)
accuracy = clf.score(Xval, Yval)
if accuracy >= best_accuracy:
print "C = ", c, " gamma = ", g, " -> ", accuracy
best_accuracy = accuracy
best_model = clf
best_c = c
best_g = g
print "Optimal SVM parameters: C = ", best_c, "and best gamma = ", best_g
print "Best validation accuracy = " + str(best_accuracy)
train_accuracy = best_model.score(Xtrain, Ytrain)
print "With training accuracy = " + str(train_accuracy)
return best_model
def build_classifier(self):
"""
build both and LDA and an SVM classifier for offline training. can lateron be used for online training
:return:
"""
self.clf = [LDA(n_components=None, priors=None, shrinkage='auto',
solver='eigen', store_covariance=False, tol=0.0001),
SVM(kernel='rbf', shrinking=True, probability=True, gamma='scale')]
# self.clf.fit(self.features, self.labels)
[c.fit(self.features, self.labels) for c in self.clf]
# possible to use methods predict(X), predict_log_proba(X) or predict_proba()
self.cv_scores = []
[self.cv_scores.append(cvs(estimator=c, X=self.features, y=self.labels, cv=10, n_jobs=-1)) for c in self.clf]
[print('mean cv score of clf {:d} is'.format(i), np.mean(cv)) for i, cv in enumerate(self.cv_scores)]
def init_SVM(params):
C = getattr(params, 'C', 1.0)
kernel = getattr(params, 'kernel', 'rbf')
degree = getattr(params, 'degree', 3)
gamma = getattr(params, 'gamma', 'auto')
coef0 = getattr(params, 'coef0', 0.0)
shrinking = getattr(params, 'shrinking', True)
probability = getattr(params, 'probability', False)
tol = getattr(params, 'tol', 0.001)
cache_size = getattr(params, 'cache_size', 200)
class_weight = getattr(params, 'class_weight', None)
verbose = getattr(params, 'verbose', False)
max_iter = getattr(params, 'max_iter', -1)
decision_function_shape = getattr(params, 'decision_function_shape', 'ovr')
random_state = getattr(params, 'random_state', None)
return SVM(C=C, kernel=kernel, degree=degree, gamma=gamma, coef0=coef0, shrinking=shrinking, probability=probability, tol=tol, cache_size=cache_size, class_weight=class_weight, verbose=verbose, max_iter=max_iter, decision_function_shape=decision_function_shape, random_state=random_state)
def impute(data, imputer, imp_method, params_dict):
imp_data = None
if imp_method == 'RandomReplace':
imp_data = imputer.replace(data, params_dict['miss_data_cond'])
elif imp_method == 'Drop':
imp_data = imputer.drop(data, params_dict['miss_data_cond'])
elif imp_method == 'Summary':
imp_data = imputer.summarize(data,
params_dict['summary_func'],
params_dict['miss_data_cond'])
elif imp_method == 'RandomForest':
clf = RandomForestClassifier(n_estimators=100, criterion='gini')
imp_data = imputer.predict(data,
params_dict['cat_cols'],
params_dict['miss_data_cond'],
clf)
elif imp_method == 'SVM':
clf = SVM()
imp_data = imputer.predict(data,
params_dict['cat_cols'],
params_dict['miss_data_cond'],
clf)
elif imp_method == 'LogisticRegression':
clf = LogisticRegression()
imp_data = imputer.predict(data,
params_dict['cat_cols'],
params_dict['miss_data_cond'],
clf)
elif imp_method == 'SVD':
imp_data = imputer.factor_analysis(data,
params_dict['cat_cols'],
params_dict['miss_data_cond'],
technique='SVD')
elif imp_method == 'KNN':
imp_data = imputer.knn(data,
params_dict['n_neighbors'],
params_dict['knn_summary_func'],
params_dict['miss_data_cond'],
params_dict['cat_cols'])
elif imp_method == 'Identity':
imp_data = data
else:
raise Exception("Imputation method {} is not valid".format(imp_method))
return imp_data
def kv_pipe(DX, labels, groups, tr_options, report_options):
'''
@pre: DX hash table database of dx histories for each patient per encounter.
DX codes are represented by index in ICD9 groups dictionary.
- labels is hash table in {s: {y: 0/1, t: time} ... } format for each subject s in subset S.
- groups is a reference hash table of indices of ICD9 group codes.
- tr_options is a hash table with options for X_training options (e.g., flat, window size, ...)
- report_options is a hash_table with options for result reporting
@post: pickle or h5 format for model and hash table of training and testing data
'''
random.seed(11)
skf = StratifiedKFold(n_splits=5, random_state=7)
splits = [(i[0], list(i[1].items())[1][1]) for i in list(labels.items())]
x_split, y_split = zip(*[(a, b) for a, b in splits])
x_split, y_split = np.array(x_split), np.array(y_split)
for train_index, test_index in skf.split(x_split, y_split):
#1. downsample training split
pos = [i for i in train_index if y_split[i] == 1]
neg = random.sample([i for i in train_index if y_split[i] == 0],
len(pos))
tr_split = x_split[neg + pos]
random.shuffle(tr_split)
te_split = x_split[test_index]
#2. split labels hash table by kfold
tr_subj = {key: labels[key] for key in tr_split}
te_subj = {key: labels[key] for key in te_split}
#3. generate train = [(x_tr,y_tr), ...] and test = [(x_te, y_te), ...]
train, test = generate_x_y(DX, tr_subj, te_subj)
x_tr, y_tr = zip(*[(a, b) for a, b in train])
x_te, y_te = zip(*[(a, b) for a, b in test])
#4. train models
lr = LR(warm_start=True, C=1e-3, penalty='l2',
verbose=1) #sag if multiclass/multilabel
svm = SVM(C=1e4,
kernel='linear',
verbose=True,
probability=True,
max_iter=1000)
rf = RF(warm_start=True, n_estimators=450, verbose=1)
return
def run_grid_search(data_x, data_y, clf_name, scoring):
if clf_name == 'lr':
print "Classifier: Logistic regression"
clfs = [LR()]
param_grid = {"C": [0.1, 0.25, 0.5, 1]}
elif clf_name == 'rf':
print "Classifier: Random forest"
clfs = [
RF(n_estimators=20,
max_depth=20,
criterion='entropy',
bootstrap=False,
max_features=20,
min_samples_split=4,
min_samples_leaf=4)
]
# use a full grid over all parameters
param_grid = {
"max_features": [20, 25],
"min_samples_split": [2, 3, 4],
"min_samples_leaf": [2, 3, 4]
}
elif clf_name == 'svm':
print "Classifier: SVM"
clfs = [SVM(penalty='l2', loss='squared_hinge', dual=False)]
param_grid = {"C": [0.001, 0.1, 1, 5]}
else:
print "Unsupported classifier:", clf_name
# run grid search
for clf in clfs:
print clf
grid_search = GridSearchCV(clf, param_grid=param_grid, scoring=scoring)
start = time()
grid_search.fit(data_x, data_y)
print(
"GridSearchCV took %.2f seconds for %d candidate parameter settings."
% (time() - start, len(grid_search.cv_results_['params'])))
report(grid_search.cv_results_)
def r_search(x, y):
#random search params
lr_params = {'penalty': ['l1', 'l2'], 'C': sp_rand(1e-5, .1)}
svm_params = {'kernel': ['rbf', 'linear'], 'C': sp_rand(10, 1e5)}
rf_params = {
'criterion': ['gini', 'entropy'],
'n_estimators': sp_randint(50, 200),
'bootstrap': [True, False]
}
gbc_params = {
'learning_rate': sp_rand(1e-6, 1e-1),
'n_estimators': sp_randint(50, 200),
'loss': ['deviance', 'exponential']
}
data = {}
xs, ys = balanced_subsample(x, y)
lst = [LR(verbose=1), RF(verbose=1), SVM(verbose=True), GBC(verbose=1)]
names = ['LR', 'RF', 'SVM', 'GB']
params = [lr_params, rf_params, svm_params, gbc_params]
for idx in range(len(lst)):
n_iter_search = 60
start = time.time()
rsearch = random_search(estimator=lst[idx],
param_distributions=params[idx],
n_iter=n_iter_search,
scoring='roc_auc',
fit_params=None,
n_jobs=1,
iid=True,
refit=True,
cv=5,
verbose=0,
random_state=8)
rsearch.fit(xs, ys)
data[names[idx]] = rsearch.cv_results_
print(names[idx] + " results complete.")
print("RandomizedSearchCV took %.2f seconds for %d candidates"
" parameter settings." % ((time.time() - start), n_iter_search))
return (data)
def opt_svm(x_train, y_train, x_valid, y_valid):
"""Obtem os melhores valores de C e kernel para o SVM"""
C = [0.01, 0.1, 1]
kernel = ['poly', 'rbf']
res = np.zeros((len(kernel), len(C)))
for i, k in enumerate(kernel):
for j, c in enumerate(C):
svm = SVM(C=c, kernel=k, random_state=0)
svm.fit(x_train, y_train)
res[i][j] = math.sqrt(sk.metrics.mean_squared_error(svm.predict(x_valid),
y_valid))
i, j = np.unravel_index(res.argmin(), res.shape)
print(f'SVM = {res[i][j]}')
print(f':: kernel = {kernel[i]}')
print(f':: C = {C[j]}')
return res[i][j]
def __init__(self, train_X, test_X, train_Y, test_Y, agent, classifier,
save_conf_mat):
self.agent = agent
if self.agent is None:
self.agent = np.ones(train_X.shape[1])
cols = np.flatnonzero(self.agent)
if (cols.shape[0] == 0):
print('[Error!] There are 0 features in the agent......')
exit(1)
self.train_X = train_X[:, cols]
self.test_X = test_X[:, cols]
self.train_Y = train_Y
self.test_Y = test_Y
self.classifier = classifier
if (self.classifier.lower() == 'knn'):
self.clf = KNN()
elif (self.classifier.lower() == 'rf'):
self.clf = RF()
elif (self.classifier.lower() == 'svm'):
self.clf = SVM()
else:
self.clf = None
print('\n[Error!] We don\'t currently support {} classifier...\n'.
format(classifier))
exit(1)
self.predictions = self.classify()
self.accuracy = self.compute_accuracy()
self.precision = self.compute_precision()
self.recall = self.compute_recall()
self.f1_score = self.compute_f1()
self.confusion_matrix = self.compute_confusion_matrix()
self.plot_confusion_matrix(save_conf_mat)
def iterate(cself, svm, classes):
cself.mention('Training SVM...')
D = spdiag(classes)
qp.update_H(D * K * D)
qp.update_Aeq(classes.T)
alphas, obj = qp.solve(cself.verbose)
# Construct SVM from solution
svm = SVM(kernel=self.kernel,
gamma=self.gamma,
p=self.p,
verbose=self.verbose,
sv_cutoff=self.sv_cutoff)
svm._X = bs.instances
svm._y = classes
svm._alphas = alphas
svm._objective = obj
svm._compute_separator(K)
svm._K = K
cself.mention('Recomputing classes...')
p_conf = svm._predictions[-bs.L_p:]
pos_classes = np.vstack([
_update_classes(part)
for part in partition(p_conf, bs.pos_groups)
])
new_classes = np.vstack(
[-np.ones((bs.L_n, 1)), pos_classes])
class_changes = round(
np.sum(np.abs(classes - new_classes) / 2))
cself.mention('Class Changes: %d' % class_changes)
if class_changes == 0:
return None, svm
return {'svm': svm, 'classes': new_classes}, None
output = net.forwardToHidden(data)
Y.append(target.to("cpu"))
X.append(output.to("cpu"))
return torch.cat(X, dim=0).data.numpy(), torch.cat(Y, dim=0).data.numpy()
print("Creating Networks")
hps = setup_hparams(sys.argv[1:])
logger, net = setup_network(hps)
net = net.to(device)
print("loading Data")
trainloader, valloader, testloader = get_dataloaders(bs=4)
print("Passing training data through network")
xtrain, ytrain = passThroughNetwork(trainloader)
# print("Passing validation data through network")
# xval, yval = passThroughNetwork(valloader)
print("Passing test data through network")
xtest, ytest = passThroughNetwork(testloader)
print("Training SVM on CPU")
svm = SVM(kernel='rbf', C=1)
svm.fit(xtrain, ytrain)
pred = svm.predict(xtest)
acc = accuracy_score(ytest, pred)
print("Accuracy on test Set:", acc * 100)
def main():
config = init()
print("Mangrove Classification")
print("SVM Classifier")
print("Classifying with: ", config.hyperparams)
print("Segmenting with: ", config.segment)
print("="*50)
if config.regenerate_features:
print("Getting data lists...")
(train_img_list, train_lbl_list), (test_img_list, test_lbl_list) = get_image_paths(config.image_paths)
print("Computing average SuperPixel size...")
avg_size = compute_avg_size(train_img_list, config.segment)
print("Average superpixel size is: {}".format(avg_size))
print("Generating features...")
train_features, train_labels = extract_features(train_img_list, train_lbl_list, config, avg_size)
test_features, test_labels = extract_features(test_img_list, test_lbl_list, config, avg_size)
print("Preprocessing...")
preprocessor = get_preprocessor(config.preprocess, train_features)
train_features = preprocessor.process(train_features)
test_features = preprocessor.process(test_features)
print("Saving features...")
train_file = open(config.save_path["train_features"], 'wb')
test_file = open(config.save_path["test_features"], 'wb')
train_save = (train_features, train_labels)
test_save = (test_features, test_labels)
pickle.dump(train_save, train_file)
pickle.dump(test_save, test_file)
train_file.close()
test_file.close()
else:
print("Loading features...")
train_file = open(config.save_path["train_features"], 'rb')
test_file = open(config.save_path["test_features"], 'rb')
train_features, train_labels = pickle.load(train_file)
test_features, test_labels = pickle.load(test_file)
train_file.close()
test_file.close()
print("Unique Labels:")
print(np.unique(train_labels))
print("Training...")
start_time = time.time()
classifier = SVM(**config.hyperparams)
classifier.fit(train_features, train_labels)
elapsed_time = time.time() - start_time
print("Training took {0:.2f} seconds".format(elapsed_time))
print("Predicting...")
start_time = time.time()
pred = classifier.predict(test_features)
elapsed_time = time.time() - start_time
print("Predicting took {0:.2f} seconds".format(elapsed_time))
report, acc, iou, precision, confusion = evaluate(test_labels, pred)
save_results(report, acc, iou, precision, confusion, config.save_path["results"])
print(report)
print()
print("Accuracy: {0:.4f}".format(acc))
print("Precision: {0:.4f}".format(precision))
print("IOU: {0:.4f}".format(iou))
print()
def test_svm(data: list, label: list) -> None:
svm = SVM(gamma="scale", kernel="linear")
svm.fit(data, label)
plot_model(np.array(data), svm)
# -*- coding: utf-8 -*-
"""
Created on Thu May 23 15:21:35 2013
@author: ed203246
"""
from sklearn import datasets
from sklearn.svm import LinearSVC as SVM
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA
from sklearn.feature_selection import SelectKBest
from epac.map_reduce.reducers import PvalPerms
import numpy
X, y = datasets.make_classification(n_samples=100,
n_features=200,
n_informative=2)
X = numpy.random.rand(*X.shape)
from epac import Perms, CV, Methods
perms_cv_svm = Perms(CV(Methods(SVM(loss="l1"), SVM(loss="l2"))), n_perms=100)
perms_cv_svm.run(X=X, y=y)
perms_cv_svm.reduce()
self = perms_cv_svm
key = 'LinearSVC(loss=l1)'
self = PvalPerms()
def svm_ai(logfile, train_features, train_labels, test_features, test_labels,
k):
return SVM(C=k, kernel='linear', cache_size=7000)
x = vect.fit_transform(x)
print('词袋处理后的第一个样本的属性为：')
print(x[0])
# 构建TF-IDF特征，归一化
tf_transformer = TfidfTransformer().fit(x)
x = tf_transformer.transform(x)
print('再经TF-IDF处理后的第一个样本的属性为：')
print(x[0])
# 随机分类训练集与测试集
trainX, testX, trainY, testY = train_test_split(
x, y, test_size=0.2, random_state=1)
# 支持向量机
modle = SVM(C=1000.0, kernel='rbf', random_state=1)
start = time.perf_counter()
modle.fit(trainX, trainY)
end = time.perf_counter()
print('Running time: %.12s ms' % ((end - start) * 1000))
# 输出评估结果
expected = testY
predict = modle.predict(testX)
print('Report:')
print(metrics.classification_report(expected, predict))
# 建立混淆矩阵
lable = list(set(expected))
matrix = pd.DataFrame(metrics.confusion_matrix(
expected, predict, labels=lable), index=lable, columns=lable)
print 'imputing with one-hot'
data_onehot = imp.binarize_data(x, cat_cols)
# replace missing data with predictions using random forest
print 'imputing with predicted values from random forest'
clf = RandomForestClassifier(n_estimators=100, criterion='gini')
data_rf = imp.predict(x, cat_cols, missing_data_cond, clf)
# replace missing data with predictions using SVM
print 'imputing with predicted values usng SVM'
clf = SVM(penalty='l2',
loss='squared_hinge',
dual=True,
tol=0.0001,
C=1.0,
multi_class='ovr',
fit_intercept=True,
intercept_scaling=1,
class_weight=None,
verbose=0,
random_state=None,
max_iter=1000)
data_svm = imp.predict(x, cat_cols, missing_data_cond, clf)
# replace missing data with predictions using logistic regression
print 'imputing with predicted values usng logistic regression'
clf = LogisticRegression(penalty='l2',
dual=False,
tol=0.0001,
C=1.0,
fit_intercept=True,
intercept_scaling=1)