def test_srgcn_cora():
args = get_default_args_for_nc("cora", "srgcn")
args.num_heads = 4
args.subheads = 1
args.nhop = 1
args.node_dropout = 0.5
args.alpha = 0.2
args.normalization = "identity"
args.attention_type = "identity"
args.activation = "linear"
norm_list = ["identity", "row_uniform", "row_softmax", "col_uniform", "symmetry"]
activation_list = ["relu", "relu6", "sigmoid", "tanh", "leaky_relu", "softplus", "elu", "linear"]
attn_list = ["node", "edge", "identity", "heat", "ppr", "gaussian"]
for norm in norm_list:
args.normalization = norm
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
args.norm = "identity"
for ac in activation_list:
args.activation = ac
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
args.activation = "relu"
for attn in attn_list:
args.attention_type = attn
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
def test_gcc_cora():
args = get_default_args_for_nc("cora", "gcc", mw="gcc_mw", dw="gcc_dw")
args.epochs = 1
args.num_workers = 0
args.batch_size = 16
args.rw_hops = 8
args.subgraph_size = 16
args.positional_embedding_size = 16
args.nce_k = 4
train(args)
def test_gcn_cora():
args = get_default_args_for_nc("cora", "gcn")
args.num_layers = 2
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
for n in ["batchnorm", "layernorm"]:
args.norm = n
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
args.residual = True
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
def __call__(self, edge_index, x=None, edge_weight=None):
if self.method_type == "emb":
if isinstance(edge_index, np.ndarray):
edge_index = torch.from_numpy(edge_index)
edge_index = (edge_index[:, 0], edge_index[:, 1])
data = Graph(edge_index=edge_index, edge_weight=edge_weight)
self.model = build_model(self.args)
embeddings = self.model(data)
elif self.method_type == "gnn":
num_nodes = edge_index.max().item() + 1
if x is None:
print("No input node features, use random features instead.")
x = np.random.randn(num_nodes, self.num_features)
if isinstance(x, np.ndarray):
x = torch.from_numpy(x).float()
if isinstance(edge_index, np.ndarray):
edge_index = torch.from_numpy(edge_index)
edge_index = (edge_index[:, 0], edge_index[:, 1])
data = Graph(x=x, edge_index=edge_index, edge_weight=edge_weight)
torch.save(data, self.data_path)
dataset = NodeDataset(path=self.data_path, scale_feat=False, metric="accuracy")
self.args.dataset = dataset
model = train(self.args)
embeddings = model.embed(data.to(model.device))
embeddings = embeddings.detach().cpu().numpy()
return embeddings
def test_dgi():
args = get_default_args_for_unsup_nn("cora", "dgi", mw="dgi_mw")
args.activation = "relu"
args.sparse = True
args.epochs = 2
ret = train(args)
assert ret["test_acc"] > 0
def test_rgcn_wn18():
args = get_default_args_kg(dataset="wn18", model="rgcn")
args.self_dropout = 0.2
args.self_loop = True
args.regularizer = "basis"
ret = train(args)
assert 0 <= ret["mrr"] <= 1
def test_attribute_mask():
args = get_default_args_generative("cora",
"gcn",
auxiliary_task="attribute_mask")
args.alpha = 1
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
def test_prone_blogcatalog():
args = get_default_args_ne(dataset="blogcatalog", model="prone")
args.step = 5
args.theta = 0.5
args.mu = 0.2
ret = train(args)
assert ret["Micro-F1 0.1"] > 0
def test_distance_to_clusters():
args = get_default_args_generative("cora",
"gcn",
auxiliary_task="distance2clusters")
args.alpha = 3
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
def test_prone_amazon():
args = get_default_args_multiplex(dataset="amazon", model="prone")
args.step = 5
args.theta = 0.5
args.mu = 0.2
ret = train(args)
assert ret["ROC_AUC"] >= 0 and ret["ROC_AUC"] <= 1
def test_prone_usa_airport():
args = get_default_args_ne(dataset="usa-airport", model="prone")
args.step = 5
args.theta = 0.5
args.mu = 0.2
ret = train(args)
assert ret["Micro-F1 0.1"] > 0
def test_pairwise_attr_sim():
args = get_default_args_generative("cora",
"gcn",
auxiliary_task="pairwise_attr_sim")
args.alpha = 100
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
def test_pairwise_distance():
args = get_default_args_generative("cora",
"gcn",
auxiliary_task="pairwise_distance")
args.alpha = 35
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
def test_sage_cora():
args = get_default_args_for_nc("cora", "sage")
args.aggr = "mean"
args.normalize = True
args.norm = "layernorm"
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
def test_netmf_ppi():
args = get_default_args_ne(dataset="ppi-ne", model="netmf")
args.window_size = 2
args.rank = 32
args.negative = 3
args.is_large = False
ret = train(args)
assert ret["Micro-F1 0.1"] > 0
def test_disengcn_cora():
args = get_default_args_for_nc("cora", "disengcn")
args.K = [4, 2]
args.activation = "leaky_relu"
args.tau = 1.0
args.iterations = 3
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
def test_clustergcn_pubmed():
args = get_default_args_for_nc("pubmed", "gcn", dw="cluster_dw")
args.cpu = True
args.batch_size = 3
args.n_cluster = 20
args.eval_step = 1
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
def test_patchy_san_mutag():
args = get_default_args_graph_clf(dataset="mutag",
model="patchy_san",
dw="patchy_san_dw")
args = add_patchy_san_args(args)
args.batch_size = 5
ret = train(args)
assert ret["test_acc"] > 0
def test_diffpool_imdb_binary():
args = get_default_args_graph_clf(dataset="imdb-b", model="diffpool")
args = add_diffpool_args(args)
args.batch_size = 100
args.train_ratio = 0.6
args.test_ratio = 0.2
ret = train(args)
assert ret["test_acc"] > 0
def test_compgcn_wn18rr():
args = get_default_args_kg(dataset="wn18rr", model="compgcn")
args.lbl_smooth = 0.1
args.score_func = "distmult"
args.regularizer = "basis"
args.opn = "sub"
ret = train(args)
assert 0 <= ret["mrr"] <= 1
def test_netsmf_ppi():
args = get_default_args_ne(dataset="ppi-ne", model="netsmf")
args.window_size = 3
args.negative = 1
args.num_round = 2
args.worker = 5
ret = train(args)
assert ret["Micro-F1 0.1"] > 0
def test_gin_mutag():
args = get_default_args_graph_clf(dataset="mutag", model="gin")
args = add_gin_args(args)
args.batch_size = 20
for kfold in [True, False]:
args.kfold = kfold
args.seed = 0
ret = train(args)
assert ret["test_acc"] > 0
def test_spectral_cora():
args = get_default_args_agc(dataset="cora",
model="prone",
mw="agc_mw",
dw="node_classification_dw")
args.model_type = "content"
args.cluster_method = "spectral"
ret = train(args)
assert ret["nmi"] >= 0
def test_mvgrl():
args = get_default_args_for_unsup_nn("cora", "mvgrl", mw="mvgrl_mw")
args.epochs = 2
args.sparse = False
args.sample_size = 200
args.batch_size = 4
args.alpha = 0.2
ret = train(args)
assert ret["test_acc"] > 0
def test_prone_ppi():
args = get_default_args_ne(dataset="ppi-ne", model="prone")
args.enhance = "prone++"
args.max_evals = 3
args.step = 5
args.theta = 0.5
args.mu = 0.2
ret = train(args)
assert ret["Micro-F1 0.1"] > 0
def test_daegc_cora():
args = get_default_args_agc(dataset="cora",
model="daegc",
mw="daegc_mw",
dw="node_classification_dw")
args.model_type = "both"
args.cluster_method = "kmeans"
ret = train(args)
assert ret["nmi"] >= 0
def test_deepwalk_wikipedia():
args = get_default_args_ne(dataset="wikipedia", model="deepwalk")
args.walk_length = 5
args.walk_num = 1
args.window_size = 3
args.worker = 5
args.iteration = 1
ret = train(args)
assert ret["Micro-F1 0.1"] > 0
def test_appnp_cora():
args = get_default_args_for_nc("cora", "ppnp")
args.num_layers = 2
args.propagation_type = "appnp"
args.alpha = 0.1
args.num_iterations = 10
ret = train(args)
assert 0 < ret["test_acc"] < 1
def test_train():
args = get_default_args(dataset="cora", model="gcn", epochs=10, cpu=True)
args.dataset = args.dataset[0]
args.model = args.model[0]
args.seed = args.seed[0]
result = train(args)
assert "test_acc" in result
assert result["test_acc"] > 0
def test_c_s_cora():
args = get_default_args_for_nc("cora", "correct_smooth_mlp")
args.use_embeddings = True
args.correct_alpha = 0.5
args.smooth_alpha = 0.5
args.num_correct_prop = 2
args.num_smooth_prop = 2
args.correct_norm = "sym"
args.smooth_norm = "sym"
args.scale = 1.0
args.autoscale = True
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
args.autoscale = False
ret = train(args)
assert 0 <= ret["test_acc"] <= 1
