dl4j: greedy layerwise trainings

I am trying to build a Deep Belief Network for dimensionality reduction. Is there a way with the dl4j extension to do greedy layerwise training on rbm layers?  I see the dl4j pretraining learner but don't understand exactly how it is supposed to be used. Are all unsupervised layers trained at the same time with this node?  Is there an example that shows how to do this?

Thanks in advance.

Hi Aaron,

the DL4J Pretraining Learner will perform greedy layerwise training on rbm layers. However, there was a bug in the DL4J training procedure for unsupervised layers which prevents it from learning anything. This bug is still present in the DL4J version we are using in the currently available KNIME version.

However, it seems to be fixed in the newest DL4J which we are using for the upcoming release but I didn't have the time to confirm this myself yet. I'll give you a note when I'm done confirming that the issue is solved.

Cheers

David