def load_hcp_tcgn(device):
time_series, labels, As = load_hcp_example()
normalized_laplacian = True
coarsening_levels = 4
graphs, perm = coarsening.coarsen(As[0],
levels=coarsening_levels,
self_connections=False)
coos = [
torch.tensor([graph.tocoo().row, graph.tocoo().col],
dtype=torch.long).to(device) for graph in graphs
]
idx_train = range(int(0.8 * time_series.shape[0]))
print('Size of train set: {}'.format(len(idx_train)))
idx_test = range(len(idx_train), time_series.shape[0])
print('Size of test set: {}'.format(len(idx_test)))
train_data = time_series[idx_train]
train_labels = labels[idx_train]
test_data = time_series[idx_test]
test_labels = labels[idx_test]
train_data = perm_data_time(train_data, perm)
test_data = perm_data_time(test_data, perm)
return graphs, coos, train_data, test_data, train_labels, test_labels
def load_hcp_tcgn(device):
time_series, labels, As = load_hcp_example(full=True)
normalized_laplacian = True
coarsening_levels = 4
shuffled = False
A = As[0]
#A = arr.todense()
if shuffled:
B = A.toarray()
B = list(B[np.triu_indices(A.shape[0])])
random.shuffle(B)
A = np.zeros((A.shape[0], A.shape[0]))
indices = np.triu_indices(A.shape[0])
A[indices] = B
A = A + A.T - np.diag(A.diagonal())
A = sp.csr_matrix(A)
graphs, perm = coarsening.coarsen(A,
levels=coarsening_levels,
self_connections=False)
#else:
W = sp.random(As[0].shape[0],
As[0].shape[0],
density=0,
format='csr',
data_rvs=lambda s: np.random.uniform(0, 1, size=s))
#graphs, perm = coarsening.coarsen(W, levels=coarsening_levels, self_connections=False)
#graphs = [As[0]]
graphs = [W]
coos = [
torch.tensor([graph.tocoo().row, graph.tocoo().col],
dtype=torch.long).to(device) for graph in graphs
]
idx_train = range(int(0.8 * time_series.shape[0]))
print('Size of train set: {}'.format(len(idx_train)))
idx_test = range(len(idx_train), time_series.shape[0])
print('Size of test set: {}'.format(len(idx_test)))
train_data = time_series[idx_train]
train_labels = labels[idx_train]
test_data = time_series[idx_test]
test_labels = labels[idx_test]
#train_data = perm_data_time(train_data, perm)
#test_data = perm_data_time(test_data, perm)
return graphs, coos, train_data, test_data, train_labels, test_labels
def experiment(args):
args.reg_weight = 5.e-4 #1.e-5
# torch.manual_seed(args.seed)
use_cuda = not args.no_cuda and torch.cuda.is_available()
device = torch.device("cuda" if use_cuda else "cpu")
#device = torch.device("cpu")
# kwargs = {'num_workers': 1, 'pin_memory': True} if use_cuda else {}
graphs, perm = create_graph(device)
coos = [
torch.tensor([graph.tocoo().row, graph.tocoo().col],
dtype=torch.long).to(device) for graph in graphs
]
train_images, test_images, train_labels, test_labels = get_mnist_data_gcn(
perm)
training_set = Dataset(train_images, train_labels)
train_loader = torch.utils.data.DataLoader(training_set,
batch_size=args.batch_size)
validation_set = Dataset(test_images, test_labels)
test_loader = torch.utils.data.DataLoader(validation_set,
batch_size=args.batch_size)
#sh = train_images.shape
#model = Net(graphs, coos)
#model = NetGCNBasic(graphs, coos)
model = NetMLP(int(graphs[0].shape[0]))
if torch.cuda.device_count() > 1:
model = torch.nn.DataParallel(model)
model.to(device)
#model.cuda()
pytorch_total_params = sum(p.numel() for p in model.parameters()
if p.requires_grad)
print(pytorch_total_params)
optimizer = optim.Adam(model.parameters(), lr=args.lr)
scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.95)
for epoch in range(1, args.epochs):
train(args, model, device, train_loader, optimizer, epoch)
scheduler.step()
test(args, model, device, test_loader, epoch)
if args.save_model:
torch.save(model.state_dict(), "mnist_cnn.pt")
def forward(self, data, graphs):
coos = [
torch.tensor(
[graph.tocoo().row, graph.tocoo().col], dtype=torch.long)
for graph in graphs
]
x, edge_index = data.x, coos[0]
try:
x = F.relu(self.conv1(x, edge_index))
except:
print("boo")
#x = F.dropout(x, training=self.training)
#x = self.conv2(x, edge_index)
x = self.fc1(x)
return F.log_softmax(x, dim=1)
def load_hcp_tcgn(device):
X_train, y_train, X_test, y_test, As = vote.load_hcp_vote()
normalized_laplacian = True
coarsening_levels = 4
graphs, perm = coarsening.coarsen(As[0],
levels=coarsening_levels,
self_connections=False)
coos = [
torch.tensor([graph.tocoo().row, graph.tocoo().col],
dtype=torch.long).to(device) for graph in graphs
]
train_data = perm_data_time(X_train, perm)
test_data = perm_data_time(X_test, perm)
return graphs, coos, train_data, test_data, y_train, y_test
