Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Getting bad results When I ran drqn experiments . #2

Open
Alex-zhai opened this issue Jun 17, 2016 · 5 comments
Open

Getting bad results When I ran drqn experiments . #2

Alex-zhai opened this issue Jun 17, 2016 · 5 comments

Comments

@Alex-zhai
Copy link

Hello, Ask you a question. When I ran exp11, the drqn agent seems not as good as dqn. I ran five games, and the drqn results are all lower than dqn. What caused this phenomenon.

@5vision
Copy link
Owner

5vision commented Jun 17, 2016

You are right. We are unable to reproduce exact results from the paper by Matthew et al. Probably, we should remove a fully connect layer before LSTM or use the same hyper-parameters, etc. We decided not to waste time on it. For all the results of DRQN in our paper, we used the implementation on Caffe.

@Alex-zhai
Copy link
Author

So, the implemention on Caffe has DRQN module?

@Alex-zhai
Copy link
Author

Can you give me some instruction of installation of DRQN on caffe? Thank you!!!

@Alex-zhai
Copy link
Author

And your environment of running DRQN on caffe is Ubantu??

@Alex-zhai
Copy link
Author

One more question, when I ran Hard-Darqn, I get cannot open <pretrain_seaquest.t7> in mode r at /tmp/luarocks_torch-scm-1-5561/torch7/lib/TH/THDiskFile.c:484. So where should I put the pretrain_seaquest.t7 model.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants