Paper
https://doi.org/10.1038/s42005-023-01500-w
Model
$$\begin{align}
\mathbf{x}(t) &= (1 - \alpha)\mathbf{x}(t - 1)
+ \alpha \tanh\!\Big(
\mathbf{g}^{\mathrm{res}}(t - 1)\, \mathbf{W}^{\mathrm{res}} \mathbf{x}(t - 1)
+ \mathbf{g}^{\mathrm{in}}(t - 1)\, \mathbf{W}^{\mathrm{in}} \mathbf{u}(t)
+ \boldsymbol{\xi}
\Big), \\\
\\\
\mathbf{g}^{\mathrm{res}}(t) &= f\!\left(
\mathbf{W}^{\mathrm{res}}_{fb} \mathbf{x}(t)
+ \mathbf{b}^{\mathrm{res}}_{fb}
\right), \\\
\\\
\mathbf{g}^{\mathrm{in}}(t) &= f\!\left(
\mathbf{W}^{\mathrm{in}}_{fb} \mathbf{x}(t)
+ \mathbf{b}^{\mathrm{in}}_{fb}
\right), \\\
\\\
\mathbf{y}(t) &= \mathbf{W}^{\mathrm{out}} \mathbf{x}(t)
+ \mathbf{b}^{\mathrm{out}}.
\end{align}$$
Official implementation
none
Notes
It would be better if the name of the model is SMESN/SMESNCell instead of the proposed SMRC in the paper.
