def create_model():
alpha = seq()
alpha.add(
Conv2D(32, [3, 3],
strides=1,
padding='valid',
activation='relu',
input_shape=(image_width, image_height, 3)))
alpha.add(Conv2D(32, [3, 3], strides=1, padding='valid',
activation='relu'))
alpha.add(MaxP2D(pool_size=(2, 2)))
alpha.add(Conv2D(64, [3, 3], strides=1, padding='valid',
activation='relu'))
alpha.add(MaxP2D(pool_size=(2, 2)))
alpha.add(Conv2D(64, [3, 3], strides=1, padding='valid',
activation='relu'))
alpha.add(
Conv2D(128, [3, 3], strides=1, padding='valid', activation='relu'))
alpha.add(MaxP2D(pool_size=(2, 2)))
alpha.add(Flatten())
alpha.add(Dense(1024, activation='relu'))
alpha.add(Dropout(0.4))
alpha.add(Dense(36, activation='softmax'))
return alpha
def compile_fit(X, Y):
#https://keras.io/guides/sequential_model/
from keras.models import Sequential as seq
from keras.layers import Dense, Activation, Dropout
from keras.optimizers import SGD
model = seq()
model.add(Dense(128, input_shape=(len(X[0]), ), activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(64, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(len(Y[0]), activation='softmax'))
#compile the model with loss, nd sochastic gradient descent
model.compile(loss='categorical_crossentropy',
optimizer=SGD(lr=0.01,
decay=1e-6,
momentum=0.9,
nesterov=True),
metrics=['accuracy'])
#convert the lists into nparrays
x = np.array(X)
y = np.array(Y)
training_board = model.fit(x,
y,
validation_split=0.25,
epochs=200,
batch_size=2,
verbose=1)
model.save('model.h5', training_board)
return training_board, model
def build_model(in_shape):
print("[log] call build_model")
model = seq()
model.add(CV2(32, 4, input_shape=in_shape))
model.add(Activation('relu'))
model.add(mp2(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(CV2(64, 3))
model.add(Activation('relu'))
model.add(CV2(64, 3))
model.add(mp2(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(512))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(nb_classes))
model.add(Activation('softmax'))
# sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
print("[log] model compile")
model.compile(loss="binary_crossentropy",
optimizer='rmsprop',
metrics=['accuracy'])
return model
from keras.datasets import mnist
(xtrain, ytrain), (xtest, ytest) = mnist.load_data()
xtrain = xtrain.reshape(60000, 28, 28, 1)
xtest = xtest.reshape(10000, 28, 28, 1)
from keras.utils import to_categorical as tc
ytrain = tc(ytrain)
ytest = tc(ytest)
from keras.models import Sequential as seq
from keras.layers import Dense, Conv2D, Flatten
model = seq()
model.add(Conv2D(2, kernel_size=3, activation="relu", input_shape=(28, 28, 1)))
i = 1
n = 4
for i in range(i):
model.add(Conv2D(filters=n, kernel_size=3, activation="relu"))
n = n * 2
model.add(Flatten())
model.add(Dense(10, activation="softmax"))
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['accuracy'])
model.fit(xtrain, ytrain, epochs=1)
pred1 = model.evaluate(xtest, ytest)
def dl(rawdata):
str = rawdata['DATE']
str_sp = []
for i1 in range(len(str)):
str_sp.append(str[i1].split('-'))
date_cov = pd.DataFrame(str_sp, columns=['year', 'month', 'day'])
rawdata_new = pd.concat([date_cov, rawdata.iloc[:, 1:]], axis=1)
id_na = rawdata_new.isnull().sum().astype('int')
ind = np.where(id_na > 0)
prcp_n = rawdata_new.iloc[:, ind[0][0]].dropna(axis=0, how='any')
hist, bin_n = np.histogram(prcp_n, density=True)
prob = hist * np.diff(bin_n)
prob.sum()
choose_bin = np.random.choice(range(len(hist)), size=1, p=prob)
l1 = np.matrix(np.where(rawdata_new.iloc[:, ind[0][0]].isnull())).T
for j1 in range(np.shape(l1)[0]):
rawdata_new.iloc[l1[j1],
ind[0][0]] = np.random.uniform(bin_n[choose_bin],
bin_n[choose_bin + 1],
size=1)
rawdata_new.iloc[:, ind[0][0]].isnull().sum()
rain_n = rawdata_new.iloc[:, ind[0][1]].dropna(axis=0, how='any')
level, freq = np.unique(rain_n, return_counts=True)
prob1 = freq / freq.sum()
l2 = np.matrix(np.where(rawdata_new.iloc[:, ind[0][1]].isnull())).T
for j2 in range(np.shape(l2)[0]):
rawdata_new.iloc[l2[j2], ind[0][1]] = np.random.choice(level,
size=1,
p=prob1)
rawdata_new.iloc[:, ind[0][1]].isnull().sum()
rawdata_new.iloc[:, ind[0][1]] = rawdata_new.iloc[:,
ind[0][1]].astype('int')
ratio = 0.7
size = rawdata_new.shape[0]
spl = int(ratio * size)
id1 = np.random.choice(range(size), spl, replace=False)
train = rawdata_new.iloc[id1, :]
test = rawdata_new.iloc[~id1, :]
id2 = rawdata_new.shape[1] - 1
train_y = train.iloc[:, id2]
test_y = test.iloc[:, id2]
train_x = train.iloc[:, range(id2)]
test_x = test.iloc[:, range(id2)]
from keras.models import Sequential as seq
from keras.layers.core import Activation, Dropout, Dense
model = seq()
model.add(
Dense(units=300,
input_dim=6,
kernel_initializer='uniform',
activation='relu'))
model.add(Dropout(0.3))
# model.add(Dense(units = 5, input_dim = 6, kernel_initializer = 'uniform', activation = 'relu'))
# model.add(Dropout(0.3))
model.add(
Dense(units=1, kernel_initializer='uniform', activation='sigmoid'))
model.compile(loss='binary_crossentropy',
optimizer='adam',
metrics=['accuracy'])
model.fit(np.array(train_x),
np.array(train_y),
epochs=50,
batch_size=80,
validation_split=0.3,
verbose=2)
# model.fit(np.array(train_x), np.array(train_y), epochs = 50, batch_size = 80, validation_split = 0.3, verbose = 1)
train_acc = model.evaluate(np.array(train_x), np.array(train_y))
test_acc = model.evaluate(np.array(test_x), np.array(test_y))
results = [train_acc[1], test_acc[1]]
return results