Add MultiNodeBatchNormalization to document

chainer · shu65 · Aug 24, 2017 · Aug 10, 2017 · Aug 10, 2017 · Aug 10, 2017
commit 568c117832947c1de26f0a214623cb092e8f9ce2
diff --git a/chainermn/links/batch_normalization.py b/chainermn/links/batch_normalization.py
@@ -13,6 +13,31 @@
 
 class MultiNodeBatchNormalization(link.Link):
 
+    """Batch normalization layer that can use the whole batch stats.
+
+    When using chainer.link.BatchNormalization, batch mean and std are
+    computed independently for the local batch in each worker. When local
+    batch size is too small, training is unstable due to unreliable batch
+    stats.
+
+    In contrast, when using this MultiNodeBatchNormalization, workers
+    communicate to conduct 'correct' batch normalization (e.g., obtaining
+    mean and std for the whole global batch).
+
+    Args:
+        size (int or tuple of ints): Size (or shape) of channel
+            dimensions.
+        comm (ChainerMN communicator): communicator to share
+            the batch stats.
+        decay (float): Decay rate of moving average. It is used on training.
+        eps (float): Epsilon value for numerical stability.
+        dtype (numpy.dtype): Type to use in computing.
+        use_gamma (bool): If ``True``, use scaling parameter. Otherwise, use
+            unit(1) which makes no effect.
+        use_beta (bool): If ``True``, use shifting parameter. Otherwise, use
+            unit(0) which makes no effect.
+    """
+
     def __init__(self, size, comm, decay=0.9, eps=2e-5, dtype=numpy.float32,
                  use_gamma=True, use_beta=True,
                  initial_gamma=None, initial_beta=None):

diff --git a/docs/source/reference/index.rst b/docs/source/reference/index.rst
@@ -15,6 +15,7 @@ Links
 
 .. autoclass:: MultiNodeChainList
     :members: add_link
+.. autoclass:: chainermn.links.MultiNodeBatchNormalization
 
 
 Optimizers and Evaluators