rbm-dbn-mnist icon indicating copy to clipboard operation
rbm-dbn-mnist copied to clipboard

Published 20 hours ago • verified publisher icon tjake

Modification for input vector of real values, not binary.

Open snurkabill opened this issue 10 years ago • 4 comments

Hi!

I would like to ask you, before I will study hard your code, if there is support for Gaussian-Bernoulli RBM. http://goo.gl/NBsyRZ part 1.4.3 ... I am newbie with RBM and I don't understand them as I need to. I have to do some project in school and your code is great source of informations. But I am not sure about real input vector.

snurkabill avatar May 06 '14 21:05 snurkabill

Yes it does there is a GaussianLayer class that does this. The minst example is pure binary though.

tjake avatar May 08 '14 16:05 tjake

can you please point to the GaussianLayer class ?

nouramoubayed avatar Mar 24 '15 13:03 nouramoubayed

https://github.com/tjake/rbm-dbn-mnist/blob/master/src/main/java/com/github/tjake/rbm/GaussianLayer.java

tjake avatar Apr 20 '15 19:04 tjake

Hi Tjake, Glad to find your project and to read your article entitled 'Resurgence in Neural Networks'. I'm having the same problem as mentioned above. I'm gonna use your codes for a real-valued dataset. As you have replied, the GaussianLayer class is supposed to do so, but in fact I can't find where and how to get use of it. I have already studied about RBM and its formulae and the differences between Gaussian/binary implementations. Would you please tell me which functions you have added(or modified) to convert your binary RBM into a Gaussian? Thanks in advance.

sinaahmadi avatar Jul 02 '15 13:07 sinaahmadi