class Client:
def __init__(self, input_shape, clients_num):
(train_images, train_labels), (test_images, test_labels) = datasets.mnist.load_data()
self.test_images = np.load("FL_QR_mnist//data//test_R.npy")
self.test_labels = test_labels
self.clients_num = clients_num
self.model = D_CNN(input_shape)
self.train_images = train_x(clients_num)
self.train_labels = train_y(clients_num)
def run_test(self, test_num):
test_images = self.test_images[:test_num]
test_labels_OneHot = to_categorical(self.test_labels[:test_num], 10)
test_images = test_images.reshape(-1, 28, 28, 1)
loss, acc = self.model.evaluate(test_images,
test_labels_OneHot, verbose=0)
return loss, acc
def train_epoch(self, client_id):
train_images_C = self.train_images[client_id]
train_labels_OneHot_C = self.train_labels[client_id]
self.model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
train_history_C = self.model.fit(train_images_C, train_labels_OneHot_C, epochs=1, batch_size=128, validation_split=0.2)
return train_history_C.history
def choose_clients(self, ratio=1.0):
choose_num = math.ceil(self.clients_num * ratio)
return np.random.permutation(self.clients_num)[:choose_num] # 序列进行随机排序
def __init__(self, input_shape, clients_num):
(train_images, train_labels), (test_images, test_labels) = datasets.mnist.load_data()
self.test_images = np.load("FL_QR_mnist//data//test_R.npy")
self.test_labels = test_labels
self.clients_num = clients_num
self.model = D_CNN(input_shape)
self.train_images = train_x(clients_num)
self.train_labels = train_y(clients_num)
def __init__(self, input_shape, clients_num):
(train_images,
train_labels), (test_images,
test_labels) = datasets.mnist.load_data()
self.test_images = test_images
self.test_labels = test_labels
self.clients_num = clients_num
self.model = D_CNN(input_shape)
self.dataset = Dataset(clients_num)
from tensorflow.keras import datasets
(train_images, train_labels), (test_images,
test_labels) = datasets.mnist.load_data()
train_images_Q = np.load(
"FL_QR_mnist//data//train_Q.npy") # "solve_DL_gradient//train_gam_32.npy"
test_images_Q = np.load(
"FL_QR_mnist//data//test_Q.npy") # "solve_DL_gradient//test_gam_32.npy"
train_images_R = np.load("FL_QR_mnist//data//train_R.npy")
test_images_R = np.load("FL_QR_mnist//data//test_R.npy")
train_labels_OneHot = to_categorical(train_labels)
test_labels_OneHot = to_categorical(test_labels)
############################# all
model = D_CNN((28, 28, 1))
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['accuracy'])
train_images = train_images[:500].reshape(-1, 28, 28, 1)
model.fit(train_images,
train_labels_OneHot[:500],
batch_size=128,
epochs=20,
validation_split=0.2)
test_images = test_images.reshape(10000, 28, 28, 1)
score = model.evaluate(test_images, test_labels_OneHot)
model.save("za_all//all_local.h5")
######################### Q_CNN ###############
model = D_CNN((28, 28, 1))