darch for dimensionality reduction

[renjithranganadhan]

I was wondering whether darch can be used for achieving dimensionality reduction. do we have any samples to refer to?

[saviola777]

Have you checked out the autoencoder example? See here for implementation details. But also check out #14, the autoencoder implementation is unfortunately still buggy.

[renjithranganadhan]

Thanks. Any fix for #14 ?

[saviola777]

Sadly, no. I don't have the knowledge to fix it (I'm not the original author) and I'm not working on this project anymore, so unless someone else takes over development of this project, I don't see a fix coming. :(

You could try to turn it into a regression problem, where input equals output, and in your network you have something like a "hour-glass shaped" network. And then after training you extract the data from one of the intermediate layers.