search

google / active-qa

Apache License 2.0
346 stars 65 forks source link

Have not found Monte Carlo Sampling in the code #21

Open fangkuann opened 5 years ago

fangkuann commented 5 years ago

Hi, Thanks for releasing the code for active-qa. After browsing the code, I did not find Monte-Carlo Sampling in the training stage. It seems that each training instance consists of only one 「query, reformulated_query, reward」 tuple. Therefore, the reward is the same for each token in one reformulated query. I don't know whether the suspicion is right. If it is right, what will model perform with or without Monte-Carlo sampling? Maybe using only one instance for Monte Carlo sampling is like the relation between stochastic gradient descent and gradient descent? Thank you

godfly commented 4 years ago

me the same, have you finger out the problem?or get a new version code? @fangkuann

fangkuann commented 4 years ago

me the same, have you finger out the problem?or get a new version code? @fangkuann

I didn't try that. But we apply the paper's training method in our query rewrite module, then the retrieve performance could be enhanced. We further improved this method by adding a value network, for details you may refer to this Chinese technic article https://www.6aiq.com/article/1577969687897.

godfly commented 4 years ago

me the same, have you finger out the problem?or get a new version code? @fangkuann

I didn't try that. But we apply the paper's training method in our query rewrite module, then the retrieve performance could be enhanced. We further improved this method by adding a value network, for details you may refer to this Chinese technic article https://www.6aiq.com/article/1577969687897.

@fangkuann Yes, I follow this article to here. May I ask some question by email? I couldn't found a way to concat you. Send a message to yanggodfly1994@gmail.com if it's ok, thanks a lot