def main():
train_dataset = SimpleDataset(total_num=1000,
is_confused=True,
x=3,
y=5,
seed=1)
test_dataset = SimpleDataset(total_num=300,
is_confused=False,
x=3,
y=5,
seed=2)
model = PassiveAggressive()
for i in range(len(train_dataset.y)):
model.fit(train_dataset.feature_vec[i], train_dataset.y[i])
print(model.w)
print(test_dataset.valid_training_result(model))
train_dataset.show_result(model.w)
random.seed(args.seed)
model = ResUNet(input_shape=(128, 128, 1),
classes=2,
filters_root=16,
depth=3)
model.summary()
if args.plot_model:
from tensorflow.python.keras.utils.vis_utils import plot_model
plot_model(model, show_shapes=True)
model.compile(loss="categorical_crossentropy",
optimizer="adam",
metrics=["categorical_accuracy"])
train_dataset = list(
zip(*list(SimpleDataset(args.train_dataset_dir_path)())))
train_dataset = (np.array(train_dataset[0]), np.array(train_dataset[1]))
x = np.array(train_dataset[0])
y = np.array(train_dataset[1])
validation_dataset = list(
zip(*list(SimpleDataset(args.validation_dataset_dir_path)())))
validation_dataset = (np.array(validation_dataset[0]),
np.array(validation_dataset[1]))
model.fit(x=x,
y=y,
validation_data=validation_dataset,
epochs=args.epochs,
batch_size=args.batch_size)
from simple_dataset import SimpleDataset
from simpleOptimizer import SimpleOptimizer
class Simple(nn.Module):
def __init__(self, indim, outdim):
#For the time being, let's work with 2 activations: relu and identity
super(Simple, self).__init__()
self.applyW = nn.Linear(indim, outdim)
def forward(self, x):
out = self.applyW(x)
return out
myDb = SimpleDataset()
dataLoader = DataLoader(myDb, batch_size=3, shuffle=True, num_workers=0)
simple = Simple(2, 1)
import torch.optim as optim
criterion = nn.MSELoss()
optimizer = optim.SGD(simple.parameters(), lr=0.001)
NUMBER_OF_EPOCHS = 20
for _ in range(NUMBER_OF_EPOCHS):
for (example_num, data) in enumerate(dataLoader):
optimizer.zero_grad()
inputs = Variable(data['example'].float())
targets = Variable(data['target'].float())
def main():
parser = argparse.ArgumentParser()
parser = argparse.ArgumentParser()
parser.add_argument('--record', type=int, default=0)
args = parser.parse_args()
is_recorded = args.record
train_dataset = SimpleDataset(total_num=250,
is_confused=True,
x=3,
y=5,
seed=1)
test_dataset = SimpleDataset(total_num=100,
is_confused=False,
x=3,
y=5,
seed=2)
model_PA = PassiveAggressive()
model_PA_one = PassiveAggressiveOne(0.05)
fig = plt.figure(figsize=(20, 4))
gs = gridspec.GridSpec(1, 10)
fig_left = fig.add_subplot(gs[0, :3])
fig_right = fig.add_subplot(gs[0, 4:])
fig_left.set_xlim([
train_dataset.dataset.x1.min() - 0.1,
train_dataset.dataset.x1.max() + 0.1
])
fig_left.set_ylim([
train_dataset.dataset.x2.min() - 0.1,
train_dataset.dataset.x2.max() + 0.1
])
fig_left.set_title("Input Data & Trained Boundary", fontsize=15)
fig_left.tick_params(labelsize=10)
fig_right.set_xlim([0, len(train_dataset.y)])
fig_right.set_ylim([0, 1])
fig_right.set_title("Test Accuracy", fontsize=15)
fig_right.tick_params(labelsize=10)
fig_right.set_xlabel("Number of training data", fontsize=12)
fig_right.set_ylabel("Accuracy", fontsize=12)
line_x = np.array(range(-10, 10, 1))
line_y = line_x * 0
line_PA, = fig_left.plot(line_x, line_y, c="#2980b9", label="PA")
line_PA_one, = fig_left.plot(line_x, line_y, c="#e74c3c", label="PA-1")
fig_left.legend(handles=[line_PA, line_PA_one], fontsize=12)
fig_right.legend(handles=[line_PA, line_PA_one], fontsize=12)
valid_result_sample = []
accuracies_PA = []
accuracies_PA_one = []
imgs = []
for i in range(len(train_dataset.y)):
fig_left.scatter(x=train_dataset.dataset.x1[i],
y=train_dataset.dataset.x2[i],
c=cm.cool(train_dataset.dataset.label[i]),
alpha=0.5)
model_PA.fit(train_dataset.feature_vec[i], train_dataset.y[i])
model_PA_one.fit(train_dataset.feature_vec[i], train_dataset.y[i])
accuracies_PA.append(test_dataset.valid_training_result(model_PA))
accuracies_PA_one.append(
test_dataset.valid_training_result(model_PA_one))
a, b, c = model_PA.w
line_y = (a * line_x + c) / (-b)
line_PA.set_data(line_x, line_y)
a, b, c = model_PA_one.w
line_y = (a * line_x + c) / (-b)
line_PA_one.set_data(line_x, line_y)
fig_right.plot(accuracies_PA, c="#2980b9", label="PA")
fig_right.plot(accuracies_PA_one, c="#e74c3c", label="PA-1")
plt.pause(0.005)
if is_recorded == 1:
matplotrecorder.save_frame()
if is_recorded == 1:
matplotrecorder.save_movie("results.mp4", 0.005)