def run(dataset,
model,
model_ssl,
runs,
epochs,
lr,
weight_decay,
early_stopping,
permute_masks=None,
logger=None):
for k in range(runs):
model.to(device).reset_parameters()
optimizer = Adam(model.parameters(), lr=lr, weight_decay=weight_decay)
if torch.cuda.is_available():
torch.cuda.synchronize()
data = dataset[0]
data = data.to(device)
num_nodes = data.num_nodes
pivot = int(num_nodes * 0.1)
cold_mask_node = range(k * pivot, (k + 1) * pivot)
data.test_mask = cold_mask_node
train_node = range(num_nodes)
train_node = [e for e in train_node if e not in cold_mask_node]
data = test_edges(data, cold_mask_node)
data.train_mask = random.sample(train_node, 140)
print("{}-fold Result".format(k))
# one possibile solution: fix the train_mask_nodes to 140 nodes
run_(data, dataset, data.total_edge_index, train_node)
data.masked_nodes = data.train_mask
train_loader = DataLoader(train_dataset, 140, shuffle=True)
data = train_edges(data, data.masked_nodes)
scheduler = StepLR(optimizer, step_size=2000, gamma=0.5)
for epoch in range(2000):
with torch.autograd.set_detect_anomaly(True):
train_loss = train_Z(model, optimizer, data, epoch)
for epoch in range(5000):
with torch.autograd.set_detect_anomaly(True):
train_loss = train(model, optimizer, data, epoch)
scheduler.step()
if torch.cuda.is_available():
torch.cuda.synchronize()
loss, acc = evaluate(model, data)
print('Val Loss: {:.4f}, Test Accuracy: {:.3f}'.format(loss, acc))
def run(dataset,
model,
runs,
epochs,
lr,
weight_decay,
early_stopping,
permute_masks=None,
logger=None):
batch_size = 30
test_losses, accs, p0, p1, r0, r1, f0, f1 = [], [], [], [], [], [], [], []
for k in range(runs):
model.to(device).reset_parameters()
optimizer = Adam(model.parameters(), lr=lr, weight_decay=weight_decay)
best_val_perf = test_perf = 0
if torch.cuda.is_available():
torch.cuda.synchronize()
# t_start = time.perf_counter()
data = dataset[0]
data = data.to(device)
num_nodes = data.num_nodes
pivot = int(num_nodes * 0.2)
cold_mask_node = range(k * pivot, (k + 1) * pivot)
data.test_mask = cold_mask_node
train_node = range(num_nodes)
train_node = [e for e in train_node if e not in cold_mask_node]
data = test_edges(data, cold_mask_node)
data.train_mask = random.sample(train_node, 140)
epoch_num = int((num_nodes - pivot) / batch_size)
print("{}-fold Result".format(k))
for i in range(5):
for epoch in range(0, epoch_num):
data.masked_nodes = train_node[epoch * batch_size:(epoch + 1) *
batch_size]
data = train_edges(data, data.masked_nodes)
train_loss = train(model, optimizer, data)
# log = 'Epoch: {:03d}, Loss: {:.4f}, Test: {:.4f}'
# print(log.format(epoch, train_loss, test_auc))
if torch.cuda.is_available():
torch.cuda.synchronize()
test_auc, test_recall, test_loss, c_m, edge_index_cs = evaluate(
model, data)
test_losses.append(test_loss)
accs.append(test_auc)
p0.append(test_recall[0][0])
p1.append(test_recall[0][1])
r0.append(test_recall[1][0])
r1.append(test_recall[1][1])
f0.append(test_recall[2][0])
f1.append(test_recall[2][1])
run_(data, dataset, train_node)
run_cs(data, dataset, edge_index_cs, train_node)
# loss, acc = tensor(train_losses), tensor(accs)
print(c_m)
print(
"Train loss: {:.4f}, Test AUC Score: {:.3f},\n precision: {:.4f}, {:.4f},\n recall: {:.4f}, {:.4f},\n fscore: {:.4f}, {:.4f}"
.format(train_loss, test_auc, test_recall[0][0], test_recall[0][1],
test_recall[1][0], test_recall[1][1], test_recall[2][0],
test_recall[2][1]))
acc_, acc, p0, p1, r0, r1, f0, f1 = tensor(test_losses), tensor(
accs), tensor(p0), tensor(p1), tensor(r0), tensor(r1), tensor(
f0), tensor(f1)
print("\n===============\n")
print("Final Result")
print(
'Loss: {:.4f}, Test AUC: {:.3f} ± {:.3f}\n precision: {:.4f}, {:.4f},\n recall: {:.4f}, {:.4f},\n fscore: {:.4f}, {:.4f}'
.format(acc_.mean().item(),
acc.mean().item(),
acc.std().item(), p0.mean(), p1.mean(), r0.mean(), r1.mean(),
f0.mean(), f1.mean()))
def run(dataset,
model,
runs,
epochs,
lr,
weight_decay,
early_stopping,
permute_masks=None,
logger=None):
batch_size = 30
losses, accs, losses_wo, accs_wo = [], [], [], []
for k in range(runs):
model.to(device).reset_parameters()
optimizer = Adam(model.parameters(), lr=lr, weight_decay=weight_decay)
best_val_perf = test_perf = 0
if torch.cuda.is_available():
torch.cuda.synchronize()
data = dataset[0]
# gdc = T.GDC(self_loop_weight=1, normalization_in='sym',
# normalization_out='col',
# diffusion_kwargs=dict(method='ppr', alpha=0.05),
# sparsification_kwargs=dict(method='topk', k=128,
# dim=0), exact=True)
# data = gdc(data)
data = data.to(device)
num_nodes = data.num_nodes
pivot = int(num_nodes * 0.1)
cold_mask_node = range(k * pivot, (k + 1) * pivot)
data.test_masked_nodes = cold_mask_node
train_node = range(num_nodes)
train_node = [e for e in train_node if e not in cold_mask_node]
data = test_edges(data, cold_mask_node)
data.train_masked_nodes = random.sample(train_node, 140)
epoch_num = int((num_nodes - pivot) / batch_size)
print("{}-fold Result".format(k))
data = train_edges(data, data.train_masked_nodes)
loss_wo, acc_wo = run_(data, dataset, data.train_edge_index,
train_node)
losses_wo.append(loss_wo)
accs_wo.append(acc_wo)
for epoch in range(1000):
with torch.autograd.set_detect_anomaly(True):
train_loss = train(model, optimizer, data, epoch)
if torch.cuda.is_available():
torch.cuda.synchronize()
loss, acc = evaluate(model, data)
losses.append(loss)
accs.append(acc)
print('Val Loss: {:.4f}, Test Accuracy: {:.3f}'.format(loss, acc))
losses, accs, losses_wo, accs_wo = tensor(losses), tensor(accs), tensor(
losses_wo), tensor(accs_wo)
print('w/o Mean Val Loss: {:.4f}, Mean Test Accuracy: {:.3f} ± {:.3f}'.
format(losses_wo.mean().item(),
accs_wo.mean().item(),
accs_wo.std().item()))
print('Mean Val Loss: {:.4f}, Mean Test Accuracy: {:.3f} ± {:.3f}'.format(
losses.mean().item(),
accs.mean().item(),
accs.std().item()))
def run(dataset,
model,
model_ssl,
runs,
epochs,
lr,
weight_decay,
early_stopping,
permute_masks=None,
logger=None):
batch_size = 30
for k in range(runs):
district = k
model.to(device).reset_parameters()
model_ssl.to(device).reset_parameters()
optimizer = Adam(model.parameters(), lr=lr, weight_decay=weight_decay)
if torch.cuda.is_available():
torch.cuda.synchronize()
data = dataset[0]
# gdc = T.GDC(self_loop_weight=1, normalization_in='sym',
# normalization_out='col',
# diffusion_kwargs=dict(method='ppr', alpha=0.05),
# sparsification_kwargs=dict(method='topk', k=128,
# dim=0), exact=True)
# data = gdc(data)
data = data.to(device)
num_nodes = data.num_nodes
pivot = int(num_nodes * 0.1)
cold_mask_node = range(k * pivot, (k + 1) * pivot)
data.test_mask = cold_mask_node
train_node = range(num_nodes)
train_node = [e for e in train_node if e not in cold_mask_node]
data = test_edges(data, cold_mask_node)
data.train_mask = random.sample(train_node, 140)
print("{}-fold Result".format(k))
# one possibile solution: fix the train_mask_nodes to 140 nodes
run_(data, dataset, data.total_edge_index, train_node)
'''
for i in range(20):
for epoch in range(0, epoch_num):
data.masked_nodes = train_node[epoch*batch_size:(epoch+1)*batch_size]
data = train_edges(data, data.masked_nodes)
with torch.autograd.set_detect_anomaly(True):
train_loss =train(model, optimizer, data)
# log = 'Epoch: {:03d}, Loss: {:.4f}, Test: {:.4f}'
# print(log.format(epoch, train_loss, test_auc))
'''
data.masked_nodes = data.train_mask
data = train_edges(data, data.masked_nodes)
scheduler = StepLR(optimizer, step_size=2000, gamma=0.5)
for epoch in range(2000):
with torch.autograd.set_detect_anomaly(True):
train_loss = train_Z(model, optimizer, data, epoch)
for epoch in range(5000):
with torch.autograd.set_detect_anomaly(True):
train_loss = train(model, optimizer, data, epoch)
scheduler.step()
if torch.cuda.is_available():
torch.cuda.synchronize()
loss, acc = evaluate(model, data)
print('Val Loss: {:.4f}, Test Accuracy: {:.3f}'.format(loss, acc))
def run(dataset, model, runs, epochs, lr, weight_decay, early_stopping,
permute_masks=None, logger=None):
batch_size = 30
losses, accs , losses_wo, accs_wo= [], [], [], []
perm = torch.randperm(dataset[0].num_nodes)
for k in range(runs):
model.to(device).reset_parameters()
optimizer = Adam(model.parameters(), lr=lr, weight_decay=weight_decay)
best_val_perf = test_perf = 0
if torch.cuda.is_available():
torch.cuda.synchronize()
writer = SummaryWriter('runs/{}_{}'.format(k, tt))
data = dataset[0]
data = data.to(device)
num_nodes = data.num_nodes
if os.path.isfile('{}_{}.pkl'.format(str(dataset)[:-2], k)):
data = pickle.load(open('{}_{}.pkl'.format(str(dataset)[:-2], k), 'rb'))
else:
pivot= int(num_nodes*0.1)
cold_mask_node = perm[list(range(k*pivot, (k+1)*pivot))]
data.test_masked_nodes =cold_mask_node
train_node = range(num_nodes)
train_node = [e for e in train_node if e not in cold_mask_node] #or unknown]
data = test_edges(data, cold_mask_node)
val_mask_node = random.sample(train_node, int(pivot*0.5))
data.val_masked_nodes = torch.tensor(val_mask_node)
data = val_edges(data, val_mask_node)
train_node = [e for e in train_node if e not in val_mask_node] #or unknown]
data.train_nodes = train_node
data.train_masked_nodes = torch.tensor(random.sample(train_node,int(num_nodes*0.1)))
data = train_edges(data, data.train_masked_nodes)
with open('{}_{}.pkl'.format(str(dataset)[:-2], k), 'wb') as f:
pickle.dump(data, f)
print("{}-fold Result".format(k))
train_node=data.train_nodes
loss_wo, acc_wo = run_(data, dataset, data.train_edge_index, train_node,writer)
losses_wo.append(loss_wo)
accs_wo.append(acc_wo)
scheduler = StepLR(optimizer, step_size=2000, gamma=0.5)
for epoch in range(2000):
with torch.autograd.set_detect_anomaly(True):
train_gan(dataset, data, writer)
for epoch in range(5000):
with torch.autograd.set_detect_anomaly(True):
train_loss =train(model, optimizer,data,epoch)
scheduler.step()
if torch.cuda.is_available():
torch.cuda.synchronize()
loss, acc = evaluate(model, data)
losses.append(loss)
accs.append(acc)
print('Val Loss: {:.4f}, Test Accuracy: {:.3f}'.format(loss,acc))
losses, accs, losses_wo, accs_wo = tensor(losses), tensor(accs), tensor(losses_wo), tensor(accs_wo)
print('w/o Mean Val Loss: {:.4f}, Mean Test Accuracy: {:.3f} ± {:.3f}'.
format(losses_wo.mean().item(),
accs_wo.mean().item(),
accs_wo.std().item()
))
print('Mean Val Loss: {:.4f}, Mean Test Accuracy: {:.3f} ± {:.3f}'.
format(losses.mean().item(),
accs.mean().item(),
accs.std().item()
))