PyG-Temporal库:用于Dynamic/Temporal图深度学习

点击上方 蓝字关注我们

PyTorch Geometric Temporal是PyTorch Geometric的动态版本。该库包含来自各种已发表研究论文的dynamic+temporal图深度学习,embedding以及spatio-temporal regression 方法。此外,它还包括一个易于使用的数据集加载器和用于dynamic and temporal graphs, gpu-support.。它还带有许多带有时间和动态图的基准数据集(您也可以创建自己的数据集)。

PyTorch Geometric Temporal使实现dynamic and temporal graphs图神经网络非常容易-。例如,这是实现a recurrent graph convolutional network with two consecutive graph convolutional GRU cells and a linear layer::

import torchimport torch.nn.functional as Ffrom torch_geometric_temporal.nn.recurrent import GConvGRUclass RecurrentGCN(torch.nn.Module):    def __init__(self, node_features, num_classes):        super(RecurrentGCN, self).__init__()        self.recurrent_1 = GConvGRU(node_features, 32, 5)        self.recurrent_2 = GConvGRU(32, 16, 5)        self.linear = torch.nn.Linear(16, num_classes)    def forward(self, x, edge_index, edge_weight):        x = self.recurrent_1(x, edge_index, edge_weight)        x = F.relu(x)        x = F.dropout(x, training=self.training)        x = self.recurrent_2(x, edge_index, edge_weight)        x = F.relu(x)        x = F.dropout(x, training=self.training)        x = self.linear(x)        return F.log_softmax(x, dim=1)

目前包含的方法有:

Discrete Recurrent Graph Convolutions

  • DCRNN from Li et al.: Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting (ICLR 2018)

  • GConvGRU from Seo et al.: Structured Sequence Modeling with Graph Convolutional Recurrent Networks (ICONIP 2018)

  • GConvLSTM from Seo et al.: Structured Sequence Modeling with Graph Convolutional Recurrent Networks (ICONIP 2018)

  • GC-LSTM from Chen et al.: GC-LSTM: Graph Convolution Embedded LSTM for Dynamic Link Prediction (CoRR 2018)

  • LRGCN from Li et al.: Predicting Path Failure In Time-Evolving Graphs (KDD 2019)

  • DyGrEncoder from Taheri et al.: Learning to Represent the Evolution of Dynamic Graphs with Recurrent Models (WWW 2019)

  • EvolveGCNH from Pareja et al.: EvolveGCN: Evolving Graph Convolutional Networks for Dynamic Graphs (AAAI 2020)

  • EvolveGCNO from Pareja et al.: EvolveGCN: Evolving Graph Convolutional Networks for Dynamic Graphs (AAAI 2020)

Temporal Graph Convolutions

  • STGCN from Yu et al.: Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting (IJCAI 2018)

Auxiliary Graph Convolutions

  • TemporalConv from Yu et al.: Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting (IJCAI 2018)

  • DConv from Li et al.: Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting (ICLR 2018)

免责声明:本文仅代表文章作者的个人观点,与本站无关。其原创性、真实性以及文中陈述文字和内容未经本站证实,请读者仅作参考,并自行核实相关内容。如发现有害或侵权内容,请联系邮箱:jubao@pinlue.com,我们将在第一时间进行核实处理。

http://image99.pinlue.com/thumb/img_jpg/ExPPKXgNVtexdPVJWGxibjBq6Uica6MibQO2uym5B94hPamvvDktjeFfobdnY0oypuTDsXYar3wZOZwatqvYoIicQQ/0.jpeg