4
$\begingroup$

I had a fairly open ended question. I've been looking at deep learning architectures for neural networks for classification in R. A few packages came up, neuralnet, H2O and nnet I've worked with all of them. nnet provides only a single hidden layer so I find that a bit limiting as it can't do deep learning. H2O is built for scale with distributed memory, but I find it takes a really long time on reasonably sized data sets (say 100s of MB), but it lets me do deep learning. neuralnet allows multiple hidden layers. If you take out the feature to do random dropouts in H2O, isn't neuralnet capable of deep learning?

$\endgroup$
1
  • 1
    $\begingroup$ broccoli - Do you have benchmarking results indicating that H2O's Deep Learning is slower than alternative CPU-based methods, for comparable amount of work (#epochs, #neurons, #hidden layers)? $\endgroup$ Jan 31 '15 at 6:58
1
$\begingroup$

a simple search on google shows R package deepnet is doing DNN :

http://cran.r-project.org/web/packages/deepnet/deepnet.pdf

$\endgroup$
-1
$\begingroup$

Now we have MXNet. But if we jump a little bit(to python), there are several nice packages like Tensorflow,Keras,Mxnet,Caffe, etc.

$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.