kinoml.ml.torch_geometric_models

Implementation of some Deep Neural Networks in Pytorch using Pytorch Geometric.

Module Contents

class kinoml.ml.torch_geometric_models.GraphConvolutionNeuralNetwork(input_shape, embedding_shape=100, hidden_shape=50, output_shape=1, activation=F.relu)

Bases: kinoml.ml.torch_models._BaseModule

Builds a Graph Convolutional Network and a feed-forward pass

Parameters

  • input_shape (int) – Number of features per node in the graph.

  • embedding_shape (int, default=100) – Dimension of latent vector.

  • hidden_shape (int, default=50) – Dimension of the hidden shape.

  • output_shape (int, default=1) – Size of the last unit, representing delta_g_over_kt in our setting.

  • _activation (torch function, default=relu) – The activation function used in the hidden (only!) layer of the network.

needs_input_shape = True
static estimate_input_shape(input_sample)
forward(data)