diff options
author | Patrick Nguyen <drpng@google.com> | 2018-04-09 16:55:06 -0700 |
---|---|---|
committer | TensorFlower Gardener <gardener@tensorflow.org> | 2018-04-09 16:57:31 -0700 |
commit | 26e36ec2c9fb061e7349b2259bc69b2140d18819 (patch) | |
tree | 9a1e2efa757def9f977c45b538e4e2c058073e8a /tensorflow/contrib/recurrent/README.md | |
parent | 66a601eece46e91c7c19cb22ebe526cf8b2253d5 (diff) |
Export recurrent and its RNN implementation in tf.contrib.
PiperOrigin-RevId: 192210794
Diffstat (limited to 'tensorflow/contrib/recurrent/README.md')
-rw-r--r-- | tensorflow/contrib/recurrent/README.md | 13 |
1 files changed, 13 insertions, 0 deletions
diff --git a/tensorflow/contrib/recurrent/README.md b/tensorflow/contrib/recurrent/README.md new file mode 100644 index 0000000000..86e10eee51 --- /dev/null +++ b/tensorflow/contrib/recurrent/README.md @@ -0,0 +1,13 @@ +# Recurrent computation library + +The recurrent computation library contains code to perform recurrent +computations. + +Its chief application is to implement recurrent neural networks (RNNs, LSTMs, +etc), which is implemented in `functional_rnn.py`. Similar techniques may be +used to implement deep networks. + +The computation saves the activations in the forward pass, and computes the +gradients in the backward pass using a single accumulator. + +The `functional_rnn` interface is compatible with the `dynamic_rnn` API. |