add rel_mask

[LittleTigger-alpha]

Q1：Hello, why add rel_mask and weight？

weight_masked = tf.nn.softmax(tf.add(rel_mask, weight), dim=0)

and according to equation 12 in paper, why not div rel_mask？
@fulifeng
Q2：dj is the number of stocks satisfying the condition sum（a_ji) > 0.
according to pagerank， d_j is the number of stocks which has relation with stock j. and i is not euqal to the i in equation 9. Is it right?