手写简单的GCN

# Simple GCN: https://towardsdatascience.com/how-to-do-deep-learning-on-graphs-with-graph-convolutional-networks-7d2250723780

import numpy as np
import torch.nn.functional

A = np.matrix([
    [0, 1, 0, 0],
    [0, 0, 1, 1],
    [0, 1, 0, 0],
    [1, 0, 1, 0]],
    dtype=float
)

I = np.matrix(np.eye(A.shape[0]))
A_hat = A + I

D_hat = np.array(np.sum(A_hat, axis=0))[0] # count the in degree as D
D_hat = np.matrix(np.diag(D_hat))

X = np.matrix([
            [i, -i]
            for i in range(A.shape[0])
        ], dtype=float)

W1 = np.matrix([
             [1, -1],
             [-1, 1]
         ])

W2 = np.matrix([
             [1],
             [-1]
         ])

X1 = D_hat**-1*A_hat*X*W1
print("X",X)
print("X1",X1)

act = torch.nn.functional.relu_(torch.tensor(X1))
print(act)