Does dnn_batch_normalization_train handle maintaining the population statistics for mean/std? Since Epsilon is a parameter for the function I'm hoping so.
On Wednesday, July 13, 2016 at 12:30:03 PM UTC-4, nouiz wrote: > > Hi, > > Just a quick email to tell about a few new features that could interrest > many people. They are merged in the development version of Theano: > http://www.deeplearning.net/software/theano/install.html#bleeding-edge-install-instructions > > 1) The wrapping of cuDNN batch norm is merged > You must use it directly and not Theano batch_normalization() function. > This should speed up execution and compilation. > > > http://deeplearning.net/software/theano_versions/dev/library/sandbox/cuda/dnn.html#batch-normalization > > 2) dilated convolution is supported on CPU and GPU > > > http://deeplearning.net/software/theano_versions/dev/library/tensor/nnet/conv.html#theano.tensor.nnet.conv2d > > 3) On the CPU, the pooling and convolution are now well multi-threaded. > > Fred > > -- --- You received this message because you are subscribed to the Google Groups "theano-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
