Closed riaanvddool closed 13 years ago
I support this - perhaps as a long lived branch. One thing we shouldn't loose sight of is getting the core methods work solid before we start to optimize.
I also want to push as much as possible a low number of dependencies for this code - which would help people just pick up and use it.
I agree, we could possibly do a sprint to speed things up if needed later, like we discussed.
What do you guys think about using Theano to bring in GPU capability?
http://deeplearning.net/software/theano/tutorial/using_gpu.html