Table 2 Fundamental steps of LSTM network in handling input Sequences.
Step | Operation | Illustration |
|---|---|---|
1 | Initialization | Initialize the hidden state \(\:{h}_{0}\) and unit state \(\:{C}_{0}\) to usually be zero vectors |
2 | Input splice | Splice the hidden state of the previous moment with the current input |
3 | Forget gate calculation | Calculate forget gate activation value using sigmoid function |
4 | Input gate calculation | Use the sigmoid function to calculate the input gate activation value |
5 | Candidate unit status | Calculate candidate unit status |
6 | Unit status update | Update cell state, combining forget and input gates |
7 | Output gate calculation | Use the sigmoid function to calculate the output gate activation value |
8 | Hide status updates | Update the hidden state as input or final output for the next time step |