def __init__(self, window_size, item_dict):
self.window_size = window_size
self.img = cv2.imread('p_map3.jpg')
self.img = cv2.resize(self.img, (64,64))
self.img = cv2.erode(self.img, np.ones((2,2)), iterations=1)
self.img = np.asarray(self.img[:,:,1] < 200 ).astype(np.uint)
#self.dot_path = []
self.current_loc = []
self.my_loc = [(120//2,64//2), 0, (-1,-1), (-1,-1)]
self.check_out_1 = Data('Check_out_1',(115//2,40//2),0,(100 // 2,25 // 2),(100 // 2,52 // 2), None)
self.check_out_2 = Data('Check_out_2',(115//2,96//2),0,(100 // 2,73 // 2),(100,100 // 2), None)
self.item_dict = item_dict
self.idx = []
self.idx_name = []
def DataDelete(self, query):
logging.error(query)
if query.tag:
datakey = Data.query(Data.tag == query.tag).get()
if query.name:
datakey = Data.query(Data.name == query.name).get()
logging.error(datakey)
if query.tag or query.name:
if datakey:
logging.error(datakey)
datakey.key.delete()
return (datakey)
def DataDelete(self,query): logging.error(query) if query.tag: datakey=Data.query(Data.tag==query.tag).get() if query.name: datakey=Data.query(Data.name==query.name).get() logging.error(datakey) if query.tag or query.name: if datakey: logging.error(datakey) datakey.key.delete() return(datakey)
n_days = 334
coefs = np.ones(n_days)
znam = np.arange(n_days + 1, 1, -1)
#print(znam.shape, coefs.shape)
coefs /= znam
coefs = np.array([coefs]).T
coefs *= coefs
def rescale_features(X):
X[:n_days, :] *= coefs
X[n_days:2 * n_days, :] *= coefs
X[2 * n_days:3 * n_days, :] *= coefs
return X
if __name__ == "__main__":
create_ds = Data('learning_set/')
ts = create_ds.create_train_set(isFull=False)
scl = StandardScaler()
X = scl.fit_transform(ts[0])
X = rescale_features(X)
cl = KMeans(n_clusters=12).fit(X)
labels = cl.predict(X)
centr = get_centriod(cl, ts[0])
book_data = acum_all_gr(create_ds, 1, ts[2])
interv = conf_intervals(book_data, labels, 0.05)
def data_fixture():
data_handler = Data()
return data_handler
def test_photo_size(self): data = Data() dataset = data.body["SN1987A"]["photometry"] types, _ = photometry.part(dataset) length = len(types) self.assertEqual(length, 5)
def test_xray_size(self): data = Data() dataset = data.body["SN1987A"]["xray"] _, flux = xray.part(dataset) length = len(flux) self.assertEqual(length, 105)
def test_spect_size(self): data = Data() dataset = data.body["SN1987A"]["spectra"] values = spectra.part(dataset) length = len(values) self.assertEqual(length, 36)
# Logistic regression using scikit learn
from sklearn.linear_model import LogisticRegression
from main import Data
# initialize data
data = Data()
data.init_data()
data.standard_split()
lr = LogisticRegression(C=100.0, random_state=1)
lr.fit(data.X_train_std, data.y_train)
plt = data.plot_decision_regions(data.X_train_std, data.y_train, lr)
plt.xlabel('petal length [standardized]')
plt.ylabel('petal length [standardized]')
plt.legend(loc='upper left')
plt.show()
import xgboost as xgb
from main import Data
import numpy as np
import pandas as pd
def predict_qlty(dt):
(X_train, Y_train, uid_train) = dt.create_train_set()
(X_test, uid_test) = dt.create_test_set()
regr = xgb.XGBRegressor(colsample_bytree=1,
subsample=0.9,
min_child_weight=5,
max_depth=6,
colsample_bylevel=0.8,
learning_rate=0.1,
n_estimators=200)
regr.fit(X_train[Y_train >= 0], Y_train[Y_train >= 0])
#print(X_test)
preds = regr.predict(X_test)
#print(preds)
preds_int = np.round(preds)
#print(preds_int)
df = pd.DataFrame({'UID': uid_test, 'score': preds_int})
#print(df)
df.to_csv('marked_data_test.csv', sep=';')
create_ds = Data('learning_set/')
predict_qlty(create_ds)
def draw_stats():
w, h = GlobalSetup.WINDOW_RESOLUTION
# draw background
Graphics.draw_rect(Graphics.shifted((0, 0)), Graphics.shifted((w, h)),
Colors.BG_COLOR)
Graphics.draw_rect(Graphics.shifted((w // 3, h // 2)),
Graphics.shifted((w * 2 // 3, h * 9 // 10)),
Colors.WHITE)
Graphics.draw_rect(Graphics.shifted((w // 3, h * 2 // 5)),
Graphics.shifted((w * 2 // 3, 0)), Colors.WHITE)
x_axis_len = w // 3 - h // 10 - 1
y_axis_len = h * 3 // 10
jump = len(Data.data) / x_axis_len
# value_jump = y_axis_len / GlobalSetup.no_balls
Graphics.plot_area(jump, y_axis_len, w, h)
# draw axes
Graphics.draw_line(Graphics.shifted((w // 3 + h // 20, h * 11 // 20)),
Graphics.shifted((w // 3 + h // 20, h * 17 // 20)))
Graphics.draw_line(
Graphics.shifted((w // 3 + h // 20, h * 11 // 20)),
Graphics.shifted((w * 2 // 3 - h // 20, h * 11 // 20)))
# draw upper arrow
Graphics.draw_line(
Graphics.shifted((w // 3 + h // 20, h * 17 // 20)),
Graphics.shifted(
(w // 3 + h // 20 - h // 200, h * 17 // 20 - h // 200)))
Graphics.draw_line(
Graphics.shifted((w // 3 + h // 20, h * 17 // 20)),
Graphics.shifted(
(w // 3 + h // 20 + h // 200, h * 17 // 20 - h // 200)))
# draw right arrow
Graphics.draw_line(
Graphics.shifted((w * 2 // 3 - h // 20, h * 11 // 20)),
Graphics.shifted(
(w * 2 // 3 - h // 20 - h // 200, h * 11 // 20 + h // 200)))
Graphics.draw_line(
Graphics.shifted((w * 2 // 3 - h // 20, h * 11 // 20)),
Graphics.shifted(
(w * 2 // 3 - h // 20 - h // 200, h * 11 // 20 - h // 200)))
shift = 25
# Labels
Graphics.write(
"Population size:",
Graphics.shifted((w // 3 + w // 30, h * 2 // 5 - 2 * shift)))
Graphics.write(
"Speed:",
Graphics.shifted((w // 3 + w // 30, h * 2 // 5 - 3 * shift)))
Graphics.write(
"Recovery:",
Graphics.shifted((w // 3 + w // 30, h * 2 // 5 - 4 * shift)))
Graphics.write(
"Social distancing:",
Graphics.shifted((w // 3 + w // 30, h * 2 // 5 - 5 * shift)))
Graphics.write(
"Ball radius:",
Graphics.shifted((w // 3 + w // 30, h * 2 // 5 - 6 * shift)))
Graphics.write(
"Infected:",
Graphics.shifted((w // 3 + w // 30, h * 2 // 5 - 8 * shift)))
Graphics.write(
"Duration:",
Graphics.shifted((w // 3 + w // 30, h * 2 // 5 - 9 * shift)))
Graphics.write(
"\u03B2 (SIR model):",
Graphics.shifted((w // 3 + w // 30, h * 2 // 5 - 10 * shift)))
Graphics.write(
"\u03B3 (SIR model):",
Graphics.shifted((w // 3 + w // 30, h * 2 // 5 - 11 * shift)))
Graphics.write(
"R0:", Graphics.shifted(
(w // 3 + w // 30, h * 2 // 5 - 12 * shift)))
# Values
Graphics.write(f"{GlobalSetup.NO_BALLS}",
Graphics.shifted((w // 2, h * 2 // 5 - 2 * shift)))
Graphics.write(f"No speed",
Graphics.shifted((w // 2, h * 2 // 5 - 3 * shift)))
Graphics.write(f"{GlobalSetup.RECOVERY_PROB} chance/iteration",
Graphics.shifted((w // 2, h * 2 // 5 - 4 * shift)))
Graphics.write(f"No social distancing",
Graphics.shifted((w // 2, h * 2 // 5 - 5 * shift)))
Graphics.write(f"{GlobalSetup.BALL_RADIUS}",
Graphics.shifted((w // 2, h * 2 // 5 - 6 * shift)))
Graphics.write(
f"{Data.data[-1][States.RECOVERED] / GlobalSetup.NO_BALLS * 100: .2f}% ({Data.data[-1][States.RECOVERED]})",
Graphics.shifted((w // 2, h * 2 // 5 - 8 * shift)))
Graphics.write(f"{len(Data.data)} iterations",
Graphics.shifted((w // 2, h * 2 // 5 - 9 * shift)))
beta, gamma, r0 = Data.SIR_analyse()
beta2, gamma2, r02 = Data.SIR_analyse_improved()
Graphics.write(f"{beta: .2f} ({beta2: .2f})",
Graphics.shifted((w // 2, h * 2 // 5 - 10 * shift)))
Graphics.write(f"{gamma: .2f} ({gamma2: .2f})",
Graphics.shifted((w // 2, h * 2 // 5 - 11 * shift)))
Graphics.write(f"{r0: .2f} ({r02: .2f})",
Graphics.shifted((w // 2, h * 2 // 5 - 12 * shift)))
tt.update()
tt.exitonclick()
def main(_):
data = Data()
# Create the model
x = tf.placeholder(tf.float32, [None, 300, 300, 1])
# Define loss and optimizer
y_ = tf.placeholder(tf.int32, [None, 116, 116])
# Build the graph for the deep net
y_conv = unet(x)
with tf.name_scope('loss'):
loss = tf.nn.sparse_softmax_cross_entropy_with_logits(labels=y_, logits=y_conv)
loss = tf.reduce_mean(loss)
tf.summary.scalar('loss', loss)
with tf.name_scope('AdamOptimizer'):
train_step = tf.train.AdamOptimizer(0.0001, 0.95, 0.99).minimize(loss)
with tf.name_scope('Accuracy'):
# Training accuracy
prediction = tf.argmax(y_conv, 3) # argmax the 3rd dimension, the label
correct_pix_pred = np.sum(prediction == y_)
incorrect_pix_pred = np.sum(prediction != y_)
n_pix = 116 * 116
train_acc = correct_pix_pred / (incorrect_pix_pred + n_pix)
# Validation accuracy
val_acc = 0
val_images = data.get_test_image_list_and_label_list()
for i in range(len(val_images[0])):
x_val = val_images[0][i]
y_val = val_images[1][i]
y_conv_acc = unet(x_val)
prediction = tf.argmax(y_conv_acc, 3)
correct_pix_pred = np.sum(prediction == y_val)
incorrect_pix_pred = np.sum(prediction != y_val)
n_pix = 116 * 116
val_acc += correct_pix_pred / (incorrect_pix_pred + n_pix)
train_acc = tf.summary.scalar('Training accuracy', train_acc)
val_acc = tf.summary.scalar('Validation accuracy', val_acc)
graph_location = tempfile.mkdtemp()
print('Saving graph to: %s' % graph_location)
writer = tf.summary.FileWriter(graph_location)
writer.add_graph(tf.get_default_graph())
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
writer = tf.summary.FileWriter(FLAGS.log_dir + '/assignment04', sess.graph)
for i in range(400):
batch = data.get_train_image_list_and_label_list()
if i % 100 == 0:
train_acc, val_acc = sess.run([train_acc, val_acc],
feed_dict={x: batch[0], y_: batch[1]})
writer.add_summary(train_acc, i)
writer.add_summary(val_acc, i)
print('step %d'% (i))
train_step.run(feed_dict={x: batch[0], y_: batch[1]})
writer.close()