Hybrid Low-Order and Higher-Order Graph Convolutional Networks
Algorithm 1
Iterative calculation for HLHG-2.
(1)
Inputs: ,, and the other parameters. N (number of hidden units), dr (dropout rate), L2 (L2 regularization), es (early stopping), epochs and lr (learning rate).
ā
Output: weight parameters and .
(2)
Randomly generate the trainable weights and ;
(3)
Iteratively calculate the forward output value(1),(2)(3);(4),(5)(6)(4)Calculate the cross entropy