def train_Amazon():
num_walks = 5
walk_length = 5
window_size = 5
neighbor_samples = 5
n_hidden = 200
n_layers = 3
activation = F.relu
dropout = 0.5
weight_decay = 0.001
lr = 0.01
epochs = 1
g, features, train_pairs, neg_neighbors, word2index, index2word = load_amazon(num_walks, walk_length, window_size,
neighbor_samples)
neg_features = []
for index, neighbors in enumerate(neg_neighbors):
neg_features.append([features[neighbor] for neighbor in neighbors])
features = torch.FloatTensor(features)
neg_features = torch.FloatTensor(neg_features)
print(features.shape)
print(neg_features.shape)
device = torch.device("cpu")
model = UN_GCN(g=g,
in_feats=features.shape[1],
n_hidden=n_hidden,
n_layers=n_layers,
activation=activation,
dropout=dropout
)
optimizer = torch.optim.Adam(model.parameters(), lr=lr, weight_decay=weight_decay)
model = model.to(device)
for epoch in range(epochs):
model.train()
h = model(features)
loss = nce_loss(h, train_pairs, neg_neighbors)
print(loss)
optimizer.zero_grad()
loss.backward()
optimizer.step()
h = model(features)
embeddings = get_embeddings(h, list(word2index))
evaluate_amazon(embeddings)
def train_amazon():
num_walks = 5
walk_length = 5
window_size = 5
neighbor_samples = 5
hidden_uints = 32
num_heads = [8]
dropout = 0.5
out_size = 200
lr = 0.1
weight_decay = 0.001
epochs = 5
g, features, train_pairs, neg_neighbors, vocab, edge_data_by_type = load_amazon(
num_walks, neighbor_samples, window_size)
features = torch.FloatTensor(features)
print("特征的shape是：", features.shape)
metapaths = []
for edge_type in edge_data_by_type.keys():
metapaths.append([edge_type] * walk_length)
model = HAN(meta_paths=metapaths,
in_size=features.shape[1],
hidden_size=hidden_uints,
out_size=out_size,
num_heads=num_heads,
dropout=dropout)
optimizer = torch.optim.Adam(model.parameters(),
lr=lr,
weight_decay=weight_decay)
for epoch in range(epochs):
model.train()
_, h = model(g, features)
loss = nce_loss(h, train_pairs, neg_neighbors)
print(loss)
optimizer.zero_grad()
loss.backward()
optimizer.step()
_, h = model(g, features)
embeddings = get_embeddings(h, list(vocab))
evaluate_amazon(embeddings)
def train_amazon():
num_walks = 5
walk_length = 5
window_size = 5
neighbor_samples = 5
weight_decay = 0.001
lr = 0.01
epochs = 1
hidden_dim = 16
out_dim = 200
num_heads = 8
g, features, train_pairs, neg_neighbors, word2index, index2word = load_amazon(
num_walks, walk_length, window_size, neighbor_samples)
features = torch.FloatTensor(features)
model = UN_GAT(g,
in_dim=features.shape[1],
hidden_dim=hidden_dim,
out_dim=out_dim,
num_heads=num_heads)
optimizer = torch.optim.Adam(model.parameters(),
lr=lr,
weight_decay=weight_decay)
device = torch.device("cpu")
model = model.to(device)
for epoch in range(epochs):
model.train()
h = model(features)
loss = nce_loss(h, train_pairs, neg_neighbors)
print(loss)
optimizer.zero_grad()
loss.backward()
optimizer.step()
h = model(features)
embeddings = get_embeddings(h, list(word2index))
evaluate_amazon(embeddings)
from metapath2vec_model import Metapath2vec
from metapath2vec_utils import load_dataset
from Linear_evaluation import evaluate_DBLP
from Amazon_evaluation import evaluate_amazon
if __name__ == "__main__":
dataset = 'Amazon'
if dataset == 'DBLP':
metapaths = [['ap', 'pa'] * 3, ['ap', 'pc', 'cp', 'pa'] * 2]
num_walk = 10
graph, sentences, node_list = load_dataset(dataset, num_walk,
metapaths)
model = Metapath2vec(sentences, graph)
model.train()
embeddings = model.get_embeddings(node_list)
evaluate_DBLP(embeddings)
if dataset == 'Amazon':
num_walk = 10
graph, sentences, index2word = load_dataset(dataset, num_walk)
model = Metapath2vec(sentences, graph)
model.train()
embeddings = model.get_embeddings(index2word)
evaluate_amazon(embeddings)