如何为一次迭代实现pagerank？

Question

我正在尝试为一次迭代实现pagerank算法。 Based on my colab here公式定义如下：

𝑟𝑗=∑𝑖→𝑗𝛽𝑟𝑖𝑑𝑖+(1−𝛽)1.𝑁

我尝试将其实现为：

r1 = (beta * (r0/degi)) + ( (1 - beta) * 1/node_count)

但是，在与networkX实现进行交叉检查时，我得到了不同的值。nx source code有点难以找到，因为它用于具有悬空值的多次迭代

我的代码（最好在colab上查看）

def one_iter_pagerank(G, beta, r0, node_id):
  # TODO: Implement this function that takes a nx.Graph, beta, r0 and node id.
  # The return value r1 is one interation PageRank value for the input node.
  # Please round r1 to 2 decimal places.

  degi = G.degree[node_id]
  node_count = G.number_of_nodes() # correct?
  r1 = (beta * (r0/degi)) + ( (1 - beta) * 1/node_count)
  print('r1:', r1)

  # crosscheck
  # alpha == beta? (without= 0.128, with=)
  r2 = nx.pagerank(G, max_iter=1, tol=0.1)[node_id]
  r3 = nx.pagerank(G, max_iter=1, tol=0.1, alpha=beta)[node_id]
  print('r2:', r2, '\nr3:', r3)


beta = 0.8
r0 = 1 / G.number_of_nodes() # assign base value?
node_id = 0
print('r0:', r0)
r1 = one_iter_pagerank(G, beta, r0, node_id)

它返回多个值：

r0: 0.029411764705882353  # base value?
r1: 0.007352941176470587  # my calculation
r2: 0.13427287581699343   # nx calc with no alpha
r3: 0.12810457516339868   # nx calc with alpha

那么，我的实现哪里错了/与nx结果有如此大的不同

colab基于Stanford CS224W course CS224W: Machine Learning with Graphs here