Weights for hidden neuron in RNN

[sai_m]

is that mandatory to maintain a constant weight (1*1 matrix) for hidden neurons in a RNN?

Thanks