gigl.src.common.models.pyg.nn.models.jumping_knowledge#
Classes
The Jumping Knowledge layer aggregation module from the "Representation Learning on Graphs with Jumping Knowledge Networks" <https://arxiv.org/abs/1806.03536> paper, which supports several aggregation schemes: concatenation ("cat"), max pooling ("max"), and weighted summation with attention scores from a bi-directional LSTM ("lstm"). |