Deep learning made easy

fastml.com

57 points·platz··8 comments

I like your article. I'd like to try.

By the way, what is the easiest way to use deep learning? Is there a hello-world example? I've been following RBMs for years but I never attempted to train one. It seemed too complex. I've used command line tools like Vowpal Wabbit with success, though.

So, is there an RBM implementation as accessible as VW?

For Matlab, there are several implementations, including the DeepLearningToolbox: https://github.com/rasmusbergpalm/DeepLearnToolbox

For Python there is the great Theano, and the RBM implementations of MORB https://github.com/benanne/morb

Is it possible to use similar algos for meshing (used in finite element analysis)? Shape recognition wouldnt be needed as the first step, since CAD files are input in FEA.

The technique looks intriguing. Are you aware of any implementation that does not require Matlab (nor its competitors)?

Theano [1] is developed at the LISA lab led by Yoshua Bengio at l'Université de Montréal.

Pylearn2 [2] is developed by the same group, and is built on top of Theano.

[1] http://deeplearning.net/software/theano/

[2] http://github.com/lisa-lab/pylearn2

Theano/Pylearn2 provide RBM, DBN, DAE, ... But do they have an implementation for sparse filtering?

Great article! I think i will try this approach on labeled EMG signals.

What is emg?

Reply on news.ycombinator.com