Recurrent Cells Reference#
This page documents all custom recurrent cells provided in the torchrecurrent.cells module.
An antisymmetric recurrent neural network cell. |
|
A gated antisymmetric recurrent neural network (RNN) cell. |
|
A Neuromodulated Bistable Recurrent cell. |
|
A Bistable recurrent cell. |
|
A Chaos Free Network (CFN) cell. |
|
A Coupled Oscillatory RNN cell. |
|
A Fast RNN cell with two scalar gates \(\alpha\) and \(\beta\). |
|
A “Fast RNN” cell with two scalar gates. |
|
An Additive–Transform Recurrent cell. |
|
An Independently Recurrent Neural Network (IndRNN) cell. |
|
A JANET (Just Another NETwork) recurrent cell. |
|
A Long Expressive Memory (LEM) recurrent cell. |
|
A Light Recurrent Unit (LightRU) cell. |
|
A Light Gated Recurrent Unit (LiGRU) cell. |
|
A Minimal Gated Unit (MGU) cell. |
|
A multiplicative LSTM cell. |
|
A Mutated Unit Type 1 (MUT1) recurrent cell. |
|
A Mutated Unit Type 2 (MUT2) recurrent cell. |
|
A Mutated Unit Type 3 (MUT3) recurrent cell. |
|
A Neural Architecture Search (NAS) cell. |
|
Original formulation of the LSTM cell (no forget gate). |
|
A Peephole LSTM cell with learnable peephole connections. |
|
A Recurrent Additive Network (RAN) cell. |
|
A Structurally Constrained Recurrent Network (SCRN) cell. |
|
A Simple Gated Recurrent Network (SGRN) cell. |
|
A Stackable Recurrent (STAR) cell. |
|
An Update Gate RNN (UGRNN) cell. |
|
An Undamped Independent Controlled Oscillatory RNN (UnICORNN) cell. |
|
A Long Short-Term Memory (LSTM) cell with working-memory connections. |