Hi, This is just to let people know that we merge PRs from Gijs van Tulder that give a much better user interface for 3d pooling and convolution and batch norm. It is the similar to 2d pooling and convolution now. You put an abstract 3d conv object and Theano later substitute it with the real implementation (cudnn, gemm, ...)
The documentation link: conv 3d: http://deeplearning.net/software/theano_versions/dev/library/tensor/nnet/conv.html#theano.tensor.nnet.conv3d pool 3d: http://deeplearning.net/software/theano_versions/dev/library/tensor/signal/pool.html#theano.tensor.signal.pool.pool_3d batch norm (grep for dnn_batch_normalization_test and dnn_batch_normalization_train): http://deeplearning.net/software/theano/library/sandbox/cuda/dnn.html Happy easier 3d/5d computation! 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.
