PyTorch Geometric Temporal supports a lot of Recurrent Graph Convolution layers. I have a static graph dataset, with ∼ 2000 timesteps and ∼ 500 nodes. Which Recurrent Graph Convolution should I try first? DCRNN, GConvGRU and GConvLSTM seem to be used in most of the examples. Which one would you try first?
PyTorch Geometric Temporal supports a lot of Recurrent Graph Convolution layers. I have a static graph dataset, with ∼ 2000 timesteps and ∼ 500 nodes. Which Recurrent Graph Convolution should I try first? DCRNN, GConvGRU and GConvLSTM seem to be used in most of the examples. Which one would you try first?