def main(opt):
train_dataset = dataloader("processedData13C_NEW.pickle", "train", 12582)
# test_dataset = bAbIDataset(opt.dataroot, opt.question_id, False)
# test_dataloader = bAbIDataloader(test_dataset, batch_size=opt.batchSize, \
# shuffle=False, num_workers=2)
opt.annotation_dim = 1 # for bAbI
#opt.n_edge_types = train_dataset.n_edge_types
#opt.n_node = train_dataset.n_node
net = GGNN(opt)
net.double()
print(net)
criterion = nn.MSELoss()
if opt.cuda:
net.cuda()
criterion.cuda()
LR = opt.lr
#optimizer = optim.Adam(net.parameters(), lr=LR )
optimizer = optim.Adam(net.parameters(), lr=LR, weight_decay=0.0015)
for epoch in range(0, opt.niter):
train(epoch, train_dataset, net, criterion, optimizer, opt)
#adjust_learning_rate(optimizer, epoch, LR)
if epoch % 1000 == 0:
test_dataset = dataloader("processedData13C_NEW.pickle", "val",
12582)
test_mse = test(test_dataset, net, criterion, optimizer, opt)
del (test_dataset)
#adjust_learning_rate(optimizer, epoch, LR, test_mse)
if test_mse < 1.75:
print("TEST_SET")
test_dataset1 = dataloader("processedData13C_NEW.pickle",
"tst", 12582)
test_mse1 = test(test_dataset1, net, criterion, optimizer, opt)
del (test_dataset1)
if test_mse < 1.73:
torch.save(net, "d:/rotation2/datasets/ggnn_net4.pkl")
print(epoch)
def __init__(self, model, loss='DiceBCE'):
df_path = '../dataset/train.csv'
root = osp.dirname(df_path)
self.num_workers = 6
self.batch_size = {'train': 3, 'val': 3}
self.accumlation_steps = 32 // self.batch_size['train']
self.lr = 5e-4
self.num_epochs = 40
self.phases = ['train', 'val']
self.device = torch.device(
'cuda:0' if torch.cuda.is_available() else 'cpu')
self.net = model.to(self.device)
self.optimizer = optim.Adam(self.net.parameters(), lr=self.lr)
self.scheduler = ReduceLROnPlateau(self.optimizer,
mode='min',
patience=3,
verbose=True)
if loss == 'BCE':
self.criterion = nn.BCELoss()
elif loss == 'Dice':
self.criterion = DiceLoss()
elif loss == 'DiceBCE':
self.criterion = DiceBCELoss()
elif loss == 'IoU':
self.criterion = IoULoss()
elif loss == 'Focal':
self.criterion = FocalLoss()
elif loss == 'Tanimoto':
self.criterion = TanimotoLoss()
else:
raise ValueError('Loss Function is not Defined')
return
self.dataloaders = {
phase: dataloader(root=root,
df_path=df_path,
phase=phase,
batch_size=self.batch_size[phase],
num_workers=self.num_workers)
for phase in self.phases
}
self.best_loss = float('inf')
self.losses = {phase: [] for phase in self.phases}
self.dice_scores = {phase: [] for phase in self.phases}
self.iou_scores = {phase: [] for phase in self.phases}
self.client = line_notify.LineNotify(
token='buNeQjYHp6sXPdwk1sMWUCmqLQr7z7czjLozKJdtevL')
avd = []
print('\n\n========================================================================')
print('==============================================================')
print('====================================================')
for round in rounds:
target=attack+'_'+classifier+'_'+feature+'_'+str(round)
excel_row=[]
# adv_samples = [f for f in listdir(adv_sample_path) if isfile(join(adv_sample_path,f))]
# for file in adv_samples:
# x_mal=np.concatenate([x_mal,np.load(adv_sample_path+file)])
# print(x_mal.shape[0])
x_ben = x_benign[0:len(x_mal)]
y_mal = np.ones(x_mal.shape[0])
y_ben = np.zeros(x_ben.shape[0])
train_data = [(x_mal, y_mal), (x_ben, y_ben)]
dataloader = dataload.dataloader(x_malware, x_benign)
# print('\n DATA SHAPE:')
print('Train --> [(x_mal:%d, y_mal:%d), (x_ben:%d, y_ben:%d)]' % (
x_mal.shape[0], y_mal.shape[0], x_ben.shape[0], y_ben.shape[0]))
# print('Test --> [(x:%d, y:%d)]' % (test_data[0].shape[0], test_data[1].shape[0]))
######EXEL######
excel_row.append(target)
excel_row.append(x_mal.shape[0])
print('\n---------------------')
print('\n EXPERIMENT ON <<' + target + '>> CLASSIFIER...')
# for arc in architechtures:
######EXEL######
# print('\n====================================')
# print('SURROGATE <<'+ str(arc)+ '>> CLASSIFIER:')
# in black box setting
classifiers = [
'SVM'
] #, 'SVM_R1', 'SVM_R2', 'SVM_R3', 'SVM_R4', 'SVM_R5', 'SVM_R6', 'SVM_R7', 'SVM_R8', 'SVM_R9', 'SVM_R10', 'SVM_R11' , 'SVM_R12', 'SVM_R13', 'SVM_R14', 'SVM_R15']# ,'RF' , 'RBF_SVM', 'LR', 'DT' , 'KNN', 'MLP']# ,'DNN']
excel_rows = []
for type in classifiers:
######EXEL######
excel_row = [type]
print('\n====================================')
print('EXPERIMENT ON <<' + type + '>> CLASSIFIER:')
x_ben = x_benign[0:len(x_mal)]
y_mal = np.ones(x_mal.shape[0])
y_ben = np.zeros(x_ben.shape[0])
train_data = [(x_mal, y_mal), (x_ben, y_ben)]
dataloader = dataload.dataloader(x_mal, x_ben)
print('\n DATA SHAPE:')
print('Train --> [(x_mal:%d, y_mal:%d), (x_ben:%d, y_ben:%d)]' %
(x_mal.shape[0], y_mal.shape[0], x_ben.shape[0], y_ben.shape[0]))
print('Test --> [(x:%d, y:%d)]' % (x_test.shape[0], y_test.shape[0]))
######EXEL######
excel_row.append(x_mal.shape[0])
#train original classifier
target_model = malware_classifeir(type=type)
target_model.train(train_data)
# ######EXEL######
# excel_row.append(accuracy_score(np.concatenate([y_mal, y_ben]),
# target_model.model.predict(np.concatenate([x_mal, x_ben]))))