Other Variants of LSTMs

Though we’ll mainly focus on the standard LSTM architecture, many variants have emerged that either simplify the complex architecture found in standard LSTMs, produce better performance, or both. We’ll look at two variants that introduce structural modifications to the cell architecture of LSTMs: peephole connections and GRUs.

Peephole connections

Peephole connections allow gates to see not only the current input and the previous final hidden state but also the previous cell state. This increases the number of weights in the LSTM cell. Having such connections has been shown to produce better results. The equations would look like these:

Let’s briefly look at how this helps the LSTM perform better. So far, the gates see the current input and final hidden state but not the cell state. However, in this configuration, if the output gate is close to zero, even when the cell state contains information crucial to better performance, the final hidden state will be close to zero. Therefore, the gates will not take the hidden state into consideration during calculation. Including the cell state directly in the gate calculation equation allows more control over the cell state, and it can perform well even in situations where the output gate is close to zero.

We illustrate the architecture of the LSTM with peephole connections in the figure below. We have grayed all the existing ...