[Doc] User Guide for Using Edge Weights in Training (#3509)

* Update

* Update

* CI

docs/source/guide/message-edge.rst

-.. _guide-message-passing-edge:
-
-2.4 Apply Edge Weight In Message Passing
-----------------------------------------
-
-:ref:`(中文版) <guide_cn-message-passing-edge>`
-
-A commonly seen practice in GNN modeling is to apply edge weight on the
-message before message aggregation, for examples, in
-`GAT <https://arxiv.org/pdf/1710.10903.pdf>`__ and some `GCN
-variants <https://arxiv.org/abs/2004.00445>`__. In DGL, the way to
-handle this is:
-
--  Save the weight as edge feature.
--  Multiply the edge feature by src node feature in message function.
-
-For example:
-
-.. code::
-
-    import dgl.function as fn
-
-    # Suppose eweight is a tensor of shape (E, *), where E is the number of edges.
-    graph.edata['a'] = eweight
-    graph.update_all(fn.u_mul_e('ft', 'a', 'm'),
-                     fn.sum('m', 'ft'))
-
-The example above uses eweight as the edge weight. The edge weight should
-usually be a scalar.
\ No newline at end of file

docs/source/guide/message.rst

@@ -33,7 +33,6 @@ The last section of it explains how to implement message passing on heterogeneou
 * :ref:`guide-message-passing-api`
 * :ref:`guide-message-passing-efficient`
 * :ref:`guide-message-passing-part`
-* :ref:`guide-message-passing-edge`
 * :ref:`guide-message-passing-heterograph`
 
 .. toctree::
@@ -44,5 +43,4 @@ The last section of it explains how to implement message passing on heterogeneou
     message-api
     message-efficient
     message-part
-    message-edge
     message-heterograph

docs/source/guide/training-eweight.rst

+.. _guide-training-eweight:
+
+5.5 Use of Edge Weights
+----------------------------------
+
+:ref:`(中文版) <guide_cn-training-eweight>`
+
+In a weighted graph, each edge is associated with a semantically meaningful scalar weight. For
+example, the edge weights can be connectivity strengths or confidence scores. Naturally, one
+may want to utilize edge weights in model development.
+
+Message Passing with Edge Weights
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+Most graph neural networks (GNNs) integrate the graph topology information in forward computation
+by and only by the message passing mechanism. A message passing operation can be viewed as
+a function that takes an adjacency matrix and additional input features as input arguments. For an
+unweighted graph, the entries in the adjacency matrix can be zero or one, where a one-valued entry
+indicates an edge. If this graph is weighted, the non-zero entries can take arbitrary scalar
+values. This is equivalent to multiplying each message by its corresponding edge weight as in
+`GAT <https://arxiv.org/pdf/1710.10903.pdf>`__.
+
+With DGL, one can achieve this by:
+
+- Saving the edge weights as an edge feature
+- Multplying the original message by the edge feature in the message function
+
+Consider the message passing example with DGL below.
+
+.. code::
+
+    import dgl.function as fn
+
+    # Suppose graph.ndata['ft'] stores the input node features
+    graph.update_all(fn.copy_u('ft', 'm'), fn.sum('m', 'ft'))
+
+One can modify it for edge weight support as follows.
+
+.. code::
+
+    import dgl.function as fn
+
+    # Save edge weights as an edge feature, which is a tensor of shape (E, *)
+    # E is the number of edges
+    graph.edata['w'] = eweight
+
+    # Suppose graph.ndata['ft'] stores the input node features
+    graph.update_all(fn.u_mul_e('ft', 'w', 'm'), fn.sum('m', 'ft'))
+
+Using NN Modules with Edge Weights
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+One can modify an NN module for edge weight support by modifying all message passing operations
+in it. The following code snippet is an example for NN module supporting edge weights.
+
+.. code::
+    import dgl.function as fn
+    import torch.nn as nn
+
+    class GNN(nn.Module):
+        def __init__(self, in_feats, out_feats):
+            super().__init__()
+            self.linear = nn.Linear(in_feats, out_feats)
+
+        def forward(self, g, feat, edge_weight=None):
+            with g.local_scope():
+                g.ndata['ft'] = self.linear(feat)
+                if edge_weight is None:
+                    msg_func = fn.copy_u('ft', 'm')
+                else:
+                    g.edata['w'] = edge_weight
+                    msg_func = fn.u_mul_e('ft', 'w', 'm')
+                g.update_all(msg_func, fn.sum('m', 'ft'))
+                return g.ndata['ft']
+
+DGL's built-in NN modules support edge weights if they take an optional :attr:`edge_weight`
+argument in the forward function.
+
+One may need to normalize raw edge weights. In this regard, DGL provides
+:func:`~dgl.nn.pytorch.conv.EdgeWeightNorm`.

docs/source/guide/training.rst

@@ -25,7 +25,7 @@ get the graph for a single-graph dataset with something like
 .. code:: python
 
     import dgl
-    
+
     dataset = dgl.data.CiteseerGraphDataset()
     graph = dataset[0]
 
@@ -54,7 +54,7 @@ The synthetic heterogeneous graph ``hetero_graph`` has these edge types:
 
     import numpy as np
     import torch
-    
+
     n_users = 1000
     n_items = 500
     n_follows = 3000
@@ -63,14 +63,14 @@ The synthetic heterogeneous graph ``hetero_graph`` has these edge types:
     n_hetero_features = 10
     n_user_classes = 5
     n_max_clicks = 10
-    
+
     follow_src = np.random.randint(0, n_users, n_follows)
     follow_dst = np.random.randint(0, n_users, n_follows)
     click_src = np.random.randint(0, n_users, n_clicks)
     click_dst = np.random.randint(0, n_items, n_clicks)
     dislike_src = np.random.randint(0, n_users, n_dislikes)
     dislike_dst = np.random.randint(0, n_items, n_dislikes)
-    
+
     hetero_graph = dgl.heterograph({
         ('user', 'follow', 'user'): (follow_src, follow_dst),
         ('user', 'followed-by', 'user'): (follow_dst, follow_src),
@@ -78,7 +78,7 @@ The synthetic heterogeneous graph ``hetero_graph`` has these edge types:
         ('item', 'clicked-by', 'user'): (click_dst, click_src),
         ('user', 'dislike', 'item'): (dislike_src, dislike_dst),
         ('item', 'disliked-by', 'user'): (dislike_dst, dislike_src)})
-    
+
     hetero_graph.nodes['user'].data['feature'] = torch.randn(n_users, n_hetero_features)
     hetero_graph.nodes['item'].data['feature'] = torch.randn(n_items, n_hetero_features)
     hetero_graph.nodes['user'].data['label'] = torch.randint(0, n_user_classes, (n_users,))
@@ -97,6 +97,7 @@ The chapter has four sections, each for one type of graph learning tasks.
 * :ref:`guide-training-edge-classification`
 * :ref:`guide-training-link-prediction`
 * :ref:`guide-training-graph-classification`
+* :ref:`guide-training-eweight`
 
 .. toctree::
     :maxdepth: 1
@@ -107,3 +108,4 @@ The chapter has four sections, each for one type of graph learning tasks.
     training-edge
     training-link
     training-graph
+    training-eweight

docs/source/guide_cn/message-edge.rst

-.. _guide_cn-message-passing-edge:
-
-2.4 在消息传递中使用边的权重
------------------------
-
-:ref:`(English Version) <guide-message-passing-edge>`
-
-一类常见的图神经网络建模的做法是在消息聚合前使用边的权重，
-比如在 `图注意力网络(GAT) <https://arxiv.org/pdf/1710.10903.pdf>`__ 和一些 `GCN的变种 <https://arxiv.org/abs/2004.00445>`__ 。
-DGL的处理方法是：
-
--  将权重存为边的特征。
--  在消息函数中用边的特征与源节点的特征相乘。
-
-例如：
-
-.. code::
-
-    import dgl.function as fn
-
-    # 假定eweight是一个形状为(E, *)的张量，E是边的数量。
-    graph.edata['a'] = eweight
-    graph.update_all(fn.u_mul_e('ft', 'a', 'm'),
-                     fn.sum('m', 'ft'))
-
-在以上代码中，eweight被用作边的权重。边权重通常是一个标量。
\ No newline at end of file

docs/source/guide_cn/message.rst

@@ -28,7 +28,6 @@
 * :ref:`guide_cn-message-passing-api`
 * :ref:`guide_cn-message-passing-efficient`
 * :ref:`guide_cn-message-passing-part`
-* :ref:`guide_cn-message-passing-edge`
 * :ref:`guide_cn-message-passing-heterograph`
 
 .. toctree::
@@ -39,5 +38,4 @@
     message-api
     message-efficient
     message-part
-    message-edge
     message-heterograph

docs/source/guide_cn/training-eweight.rst

+.. _guide_cn-training-eweight:
+
+5.5 使用边权重
+----------------------------------
+
+:ref:`(English Version) <guide-training-eweight>`
+
+在一个加权图里，每条边都有一个有意义的标量权重。例如，边权重可以是连接强度或者信心指数。
+人们自然会想要在模型开发中使用它们。
+
+使用边权重的消息传递
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+大部分图神经网络在前馈计算中仅通过消息传递引入图结构信息。一个消息传递运算可以视为一个函数。
+这个函数的输入变量是一个邻接矩阵和其他输入特征。对于一个不带权重的图，邻接矩阵里的元素不是零就是一。
+值为一的元素表示一条边。对于一个加权图，非零的元素可以取任意标量值。这等价于把每条消息和对应的边权重相乘，
+即`图注意力网络 <https://arxiv.org/pdf/1710.10903.pdf>`__中的做法。
+
+在DGL里可以通过以下步骤实现这一需求：
+
+- 把边权重保存为一个边特征
+- 在消息函数里，用保存的边特征与对应边的原始消息相乘
+
+考虑以下基于DGL的消息传递示例：
+
+.. code::
+
+    import dgl.function as fn
+
+    # 假定graph.ndata['ft']存储了输入节点特征
+    graph.update_all(fn.copy_u('ft', 'm'), fn.sum('m', 'ft'))
+
+可以将其按以下方式修改以支持边权重：
+
+.. code::
+
+    import dgl.function as fn
+
+    # 将边权重保存为一个边特征。边权重是一个形状为(E, *)的张量。
+    # E是边的数量
+    graph.edata['w'] = eweight
+
+    # 假定graph.ndata['ft']存储了输入节点特征
+    graph.update_all(fn.u_mul_e('ft', 'w', 'm'), fn.sum('m', 'ft'))
+
+在NN模块中使用边权重
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+用户可以通过修改NN模块中所有的消息传递操作来给NN模块增加边权重支持。以下代码块提供了一个例子。
+
+.. code::
+    import dgl.function as fn
+    import torch.nn as nn
+
+    class GNN(nn.Module):
+        def __init__(self, in_feats, out_feats):
+            super().__init__()
+            self.linear = nn.Linear(in_feats, out_feats)
+
+        def forward(self, g, feat, edge_weight=None):
+            with g.local_scope():
+                g.ndata['ft'] = self.linear(feat)
+                if edge_weight is None:
+                    msg_func = fn.copy_u('ft', 'm')
+                else:
+                    g.edata['w'] = edge_weight
+                    msg_func = fn.u_mul_e('ft', 'w', 'm')
+                g.update_all(msg_func, fn.sum('m', 'ft'))
+                return g.ndata['ft']
+
+DGL内置的NN模块如果在forward函数中支持一个可选的:attr:`edge_weight`变量，那么它们已经支持了边权重。
+
+用户可能会需要标准化原始边权重。DGL提供了一个满足这个功能的函数
+:func:`~dgl.nn.pytorch.conv.EdgeWeightNorm`。

docs/source/guide_cn/training.rst

@@ -20,7 +20,7 @@
 .. code:: python
 
     import dgl
-    
+
     dataset = dgl.data.CiteseerGraphDataset()
     graph = dataset[0]
 
@@ -46,7 +46,7 @@
 
     import numpy as np
     import torch
-    
+
     n_users = 1000
     n_items = 500
     n_follows = 3000
@@ -55,14 +55,14 @@
     n_hetero_features = 10
     n_user_classes = 5
     n_max_clicks = 10
-    
+
     follow_src = np.random.randint(0, n_users, n_follows)
     follow_dst = np.random.randint(0, n_users, n_follows)
     click_src = np.random.randint(0, n_users, n_clicks)
     click_dst = np.random.randint(0, n_items, n_clicks)
     dislike_src = np.random.randint(0, n_users, n_dislikes)
     dislike_dst = np.random.randint(0, n_items, n_dislikes)
-    
+
     hetero_graph = dgl.heterograph({
         ('user', 'follow', 'user'): (follow_src, follow_dst),
         ('user', 'followed-by', 'user'): (follow_dst, follow_src),
@@ -70,7 +70,7 @@
         ('item', 'clicked-by', 'user'): (click_dst, click_src),
         ('user', 'dislike', 'item'): (dislike_src, dislike_dst),
         ('item', 'disliked-by', 'user'): (dislike_dst, dislike_src)})
-    
+
     hetero_graph.nodes['user'].data['feature'] = torch.randn(n_users, n_hetero_features)
     hetero_graph.nodes['item'].data['feature'] = torch.randn(n_items, n_hetero_features)
     hetero_graph.nodes['user'].data['label'] = torch.randint(0, n_user_classes, (n_users,))
@@ -88,6 +88,7 @@
 * :ref:`guide_cn-training-edge-classification`
 * :ref:`guide_cn-training-link-prediction`
 * :ref:`guide_cn-training-graph-classification`
+* :ref:`guide_cn-training-graph-eweight`
 
 .. toctree::
     :maxdepth: 1