Recurrent Layers¶
Recurrent neural network layers are provided in the following two modules:
mxnet.gluon.rnn and mxnet.gluon.contrib.rnn.
Recurrent Cells¶
rnn.LSTMCell(hidden_size[, …]) |
Long-Short Term Memory (LSTM) network cell. |
rnn.GRUCell(hidden_size[, …]) |
Gated Rectified Unit (GRU) network cell. |
rnn.RecurrentCell([prefix, params]) |
Abstract base class for RNN cells |
rnn.SequentialRNNCell([prefix, params]) |
Sequentially stacking multiple RNN cells. |
rnn.BidirectionalCell(l_cell, r_cell[, …]) |
Bidirectional RNN cell. |
rnn.DropoutCell(rate[, axes, prefix, params]) |
Applies dropout on input. |
rnn.ZoneoutCell(base_cell[, …]) |
Applies Zoneout on base cell. |
rnn.ResidualCell(base_cell) |
Adds residual connection as described in Wu et al, 2016 (https://arxiv.org/abs/1609.08144). |
contrib.rnn.Conv1DRNNCell(input_shape, …) |
1D Convolutional RNN cell. |
contrib.rnn.Conv2DRNNCell(input_shape, …) |
2D Convolutional RNN cell. |
contrib.rnn.Conv3DRNNCell(input_shape, …) |
3D Convolutional RNN cells |
contrib.rnn.Conv1DLSTMCell(input_shape, …) |
1D Convolutional LSTM network cell. |
contrib.rnn.Conv2DLSTMCell(input_shape, …) |
2D Convolutional LSTM network cell. |
contrib.rnn.Conv3DLSTMCell(input_shape, …) |
3D Convolutional LSTM network cell. |
contrib.rnn.Conv1DGRUCell(input_shape, …) |
1D Convolutional Gated Rectified Unit (GRU) network cell. |
contrib.rnn.Conv2DGRUCell(input_shape, …) |
2D Convolutional Gated Rectified Unit (GRU) network cell. |
contrib.rnn.Conv3DGRUCell(input_shape, …) |
3D Convolutional Gated Rectified Unit (GRU) network cell. |
contrib.rnn.VariationalDropoutCell(base_cell) |
Applies Variational Dropout on base cell. |
contrib.rnn.LSTMPCell(hidden_size, …[, …]) |
Long-Short Term Memory Projected (LSTMP) network cell. |
Recurrent Layers¶
rnn.RNN(hidden_size[, num_layers, …]) |
Applies a multi-layer Elman RNN with tanh or ReLU non-linearity to an input sequence. |
rnn.LSTM(hidden_size[, num_layers, layout, …]) |
Applies a multi-layer long short-term memory (LSTM) RNN to an input sequence. |
rnn.GRU(hidden_size[, num_layers, layout, …]) |
Applies a multi-layer gated recurrent unit (GRU) RNN to an input sequence. |