Deep Belief Network Implementation in Java?

Hello all,

I'm looking for a decent implementation of deep belief networks in Java. I've found jaRBM, but I think that it is just a RBM implementation and it seems not being updated anymore (Although I haven't fully checked it). I know there are stuffs written in python, but I need a Java implementation.

asked Aug 05 '10 at 08:33

cglr
195●4●8●11

edited Aug 05 '10 at 08:53

ogrisel
4989●9●55●91

3 Answers:

oldest newest most voted

0	If you have a c/c++ implementation, you can make a wrapper to it and call it from java. answered Aug 09 '10 at 11:20 Nouiz 61●2 No, unfortunately not, if I had a C/C++ implementation, I'm aware that it is possible to call these libs or APIs with JNI. (Aug 10 '10 at 03:16) cglr

If you have RBMs or stacked RBMs, it is trivial to get DBNs from that, as far as sampling from the DBN is concerned. If you want to fine-tune the model for something like supervised prediction or classification, you also need neural net code, but that already exist in Java.

answered Aug 09 '10 at 11:20

Yoshua Bengio
481●1●5●7

I know that the Mahout project is planning to add an implementation of stacked RBMs. Apparently the work has started in this gihub branch but I haven't tried it yet.

answered Aug 05 '10 at 08:57

ogrisel
4989●9●55●91

edited Aug 05 '10 at 08:58

hmm great news ogriel. i'm going to review the code and try it if it is usable or not.

(Aug 05 '10 at 09:10) cglr

Subscription:

Once you sign in you will be able to subscribe for any updates here

Tags:

deep-learning ×128
rbm ×121
deep-belief-networks ×33
java ×31

Asked: Aug 05 '10 at 08:33

Seen: 4,714 times

Last updated: Aug 10 '10 at 03:16

Deep Belief Network Implementation in Java?

Subscription:

Related questions