Vanilla Recurrent Neural Network

Hi all!

I am designing a neural network and lacking the functionality for creating a simple recurrent layer that can be represented by a fully-connected layer, which outputs are to be fed back to input [I attached the figure]. The LSTM and GRU do not work in my case, as I need to implement only the multiplication of the inputs and then fed the results back. I also need this network to have three hidden layers (as it is shown on the picture)—the existing Neural Net Time Series tool, unfortunately, does not work for me either.

Is there any way to design such a network without receiving an error message ["Error: Connection cycle."]?

Thank you in advance!