# dgl.softmax_nodes¶

dgl.softmax_nodes(graph, feat, *, ntype=None)[source]

Perform graph-wise softmax on the node features.

For each node $$v\in\mathcal{V}$$ and its feature $$x_v$$, calculate its normalized feature as follows:

$z_v = \frac{\exp(x_v)}{\sum_{u\in\mathcal{V}}\exp(x_u)}$

If the graph is a batch of multiple graphs, each graph computes softmax independently. The result tensor has the same shape as the original node feature.

Parameters
• graph (DGLGraph.) – The input graph.

• feat (str) – The node feature name.

• ntype (str, optional) – The node type name. Can be omitted if there is only one node type in the graph.

Returns

Result tensor.

Return type

Tensor

Examples

>>> import dgl
>>> import torch as th


Create two DGLGraph objects and initialize their node features.

>>> g1 = dgl.graph(([0, 1], [1, 0]))              # Graph 1
>>> g1.ndata['h'] = th.tensor([1., 1.])
>>> g2 = dgl.graph(([0, 1], [1, 2]))              # Graph 2
>>> g2.ndata['h'] = th.tensor([1., 1., 1.])


Softmax over one graph:

>>> dgl.softmax_nodes(g1, 'h')
tensor([.5000, .5000])


Softmax over a batched graph:

>>> bg = dgl.batch([g1, g2])
>>> dgl.softmax_nodes(bg, 'h')
tensor([.5000, .5000, .3333, .3333, .3333])