thudm / mcns Goto Github PK

hi authors,
this work is difficult to me to understand, do you have slides or video including more details ?

Is there also example code for the node classification task as reported in the paper?

Thank you.

There was a define about central node in figure 2 said "The central nodes are traversed by DFS", but I can't see define of node1 by dfs in code, and only see dfs used to sample next node in MH sampler.
Looking forward to your reply, please point out if there is any problem with my understanding

Hi,

Is there an example of the link prediction task?

Hello,

I am currently studying negative sampling in the recommendation systems field. In your paper, you compared the current negative sampling methods and your proposed method. In your repo, I could not see the state-of-the-art methods' functions. How can I access the state-of-the-art methods' codes that you mentioned in your paper? Can you help me with this?

Thanks in advance.

请教一下, 在 deepwalk 推荐任务代码实现中, 计算 loss 或 accuracy 时, 为什么负样本采样个数为 1 呢?
实际中不是应该设置为 k 么

def _loss(self):
    aff = tf.reduce_sum(tf.multiply(self.outputs1, self.outputs2), 1) + self.true_b
    neg_aff = tf.reduce_sum(tf.multiply(self.outputs1, self.neg_outputs), 1) + self.neg_b
    # xent_loss
    true_xent = tf.nn.sigmoid_cross_entropy_with_logits(labels=tf.ones_like(aff), logits=aff)
    negative_xent = tf.nn.sigmoid_cross_entropy_with_logits(labels=tf.zeros_like(neg_aff), logits=neg_aff)
    loss = tf.reduce_sum(true_xent) + tf.reduce_sum(negative_xent)

    self.loss = loss / tf.cast(self.batch_size, tf.float32)
    self.merged_loss = tf.summary.scalar('merged_loss', self.loss)

论文中也是设置为 k 的.

Hi，author
Whether the argument, 'user_num' and 'item_num', in recommendation need to be pre-defined？