I think the way you are doing backpropagation through time in the Elmann RNN is not correct. You are updating the weights at every step of the sequence, in this way you are not doing BPTT. I am working on a similar problem and I ended up in your blog.
I think the way you are doing backpropagation through time in the Elmann RNN is not correct. You are updating the weights at every step of the sequence, in this way you are not doing BPTT. I am working on a similar problem and I ended up in your blog.