def run_method_2():
# Replicates the method outlined in the Google Doc
source = data_source.DataSource()
T = 100
K = 10
zero_vec = np.zeros(10)
ws = [zero_vec, zero_vec]
gs = [zero_vec]
for t in range(T):
w_cur = ws[-1]
w_prev = ws[-2]
g_prev = gs[-1]
# Generate K random machines to perform the modified DSVRG updates
inner_results = [
machine.Machine(source, data_count=100).execute_modified_DSVRG(
w_cur, w_prev, g_prev) for _ in range(K)
]
w_next = np.mean([w for (w, _) in inner_results])
g_cur = np.mean([g for (_, g) in inner_results])
ws.append(w_next)
gs.append(g_cur)
pass
def loadData(self): r = data.DataSource() def update(r): print r r.loadData(update, const.COLUMN_ID, const.COLUMN_DATE, const.COLUMN_VIDEOSID, videoSid = 'tvn8opw0c3pq')
lw=2) for cen in centroids
])
tags = ['x', 'y'] + [title for (_, title) in hm.images]
show_many_images(
imgs,
tags,
patchess,
)
c = tf.ConfigProto()
c.gpu_options.allow_growth = True
sess = tf.Session(config=c)
keras.backend.tensorflow_backend.set_session(sess)
ds = data_source.DataSource(PREFIX)
m = model.Model(os.path.join(PREFIX, sys.argv[1]), ds)
# print(m.eval_accuracies())
for i, name in enumerate(TEST_NAMES):
# exit()
# exit()
# for name, x, y, i in zip(ds.names, ds.xtrain, ds.ytrain, range(100000)):
# n = ds.names
# np.random.shuffle(n)
# for i, name in enumerate(n):
# for i, name in enumerate([
# '17-10-24-23-10-39_blue-thunderbluff-courtyard-scroll0_marilyn',
# '17-10-24-23-28-52_blue-darnassus-auctionhouse-scroll0_gina',
# '17-10-24-23-36-13_blue-darnassus-temple-scroll0_kelly-occlusion',
# '17-10-28-21-24-48_red-stonetalon-sunrock-scroll10_anna',
self.model.fit(train_data, label_data)
def predict(self, test_data, test_label):
expected = test_label
predicted = self.model.predict(test_data)
predicted_prob = self.model.predict_proba(test_data)
print "predicted_prob"
print predicted_prob
print "==========================="
print(metrics.classification_report(expected, predicted))
#print(metrics.confusion_matrix(expected, predicted))
print(metrics.roc_auc_score(expected, predicted))
print self.model.coef_[0]
if __name__ == "__main__":
file_name = "ccf_offline_stage1_train.csv"
data = data_source.DataSource()
(features, labels) = data.load_normalize_data(file_name)
print features
print labels
#sys.exit()
(X_train, X_test, y_train,
y_test) = cross_validation.train_test_split(features,
labels,
test_size=0.3)
model = Model()
model.train(X_train, y_train)
model.predict(X_test, y_test)