epochs = 10000 # Number of training epochs
es_patience = 100 # Patience fot early stopping
# Preprocessing operations
X = preprocess_features(X)
A = A + np.eye(A.shape[0]) # Add self-loops
# Model definition (as per Section 3.3 of the paper)
X_in = Input(shape=(F,))
A_in = Input(shape=(N,))
dropout1 = Dropout(dropout_rate)(X_in)
graph_attention_1 = GraphAttention(F_,
attn_heads=n_attn_heads,
attn_heads_reduction='concat',
dropout_rate=dropout_rate,
activation='elu',
kernel_regularizer=l2(l2_reg),
attn_kernel_regularizer=l2(l2_reg))([dropout1, A_in])
dropout2 = Dropout(dropout_rate)(graph_attention_1)
graph_attention_2 = GraphAttention(n_classes,
attn_heads=1,
attn_heads_reduction='average',
dropout_rate=dropout_rate,
activation='softmax',
kernel_regularizer=l2(l2_reg),
attn_kernel_regularizer=l2(l2_reg))([dropout2, A_in])
# Build model
model = Model(inputs=[X_in, A_in], outputs=graph_attention_2)
optimizer = Adam(lr=learning_rate)
learning_rate = 0.005 # Learning rate for SGD
epochs = 2000 # Number of epochs to run for
es_patience = 100
# Preprocessing operations
X_train /= X_train.sum(1).reshape(-1, 1)
A_train = A_train.toarray()
# Model definition (as per Section 3.3 of the paper)
X = Input(shape=(F, ))
A = Input(shape=(N, ))
dropout1 = Dropout(dropout_rate)(X)
graph_attention_1 = GraphAttention(
F_,
attn_heads=8,
attn_heads_reduction='concat',
activation='elu',
kernel_regularizer=l2(l2_reg))([dropout1, A])
dropout2 = Dropout(dropout_rate)(graph_attention_1)
graph_attention_2 = GraphAttention(
n_classes,
attn_heads=1,
attn_heads_reduction='average',
activation='softmax',
kernel_regularizer=l2(0.0005))([dropout2, A])
# Build model
model = Model(inputs=[X, A], outputs=graph_attention_2)
optimizer = Adam(lr=0.005)
model.compile(optimizer=optimizer,
loss='categorical_crossentropy',