Layers

Please first read the class naming parlance section first.

The set of layers, layer settings and layer factories included with this package are listed below:

• Convolution:

  • ConvolutionLayerFactory: Create convolution layers.

  • PoolFactory: Create a 2D max pool and output it’s shape.

  • MaxPool2dFactory: Create a 2D max pool and output it’s shape.

  • MaxPool1dFactory: Create a 1D max pool and output it’s shape.

• Linear:

  • DeepLinearNetworkSettings: Settings for a deep fully connected network.

  • DeepLinear: A layer that has contains one more nested layers, including batch normalization and activation.

• Recurrent (RNN, LSTM, GRU):

  • RecurrentAggregationNetworkSettings: Settings for a recurrent neural network.

  • RecurrentAggregation: A recurrent neural network model with an output aggregation.

• Conditional Random Fields:

  • CRF: Conditional random field (pure PyTorch nn.Module).

  • RecurrentCRFNetworkSettings: Settings for a recurrent neural network.

  • RecurrentCRF: Adapt a CRF module using the framework based BaseNetworkModule class.